Construction method and application of pre-eclampsia risk assessment and prediction model

Abstract

The invention provides a construction method of a pre-eclampsia risk assessment prediction model, a prediction product, a prediction system and related applications thereof. Specifically, a logistic regression analysis model is constructed based on obtained clinical indexes of a pregnant woman, and the prior probability of premature eclampsia or late eclampsia of the pregnant woman is obtained; carrying out median multiple calibration on the obtained MAP and IGFBP1 and / or PLGF of the pregnant woman, and respectively constructing multivariate normal distribution models of the sick pregnant woman and the healthy pregnant woman based on the MAP MoM value and the IGFBP1 MoM value and / or the PLGF MoM value to obtain the conditional probability of the premature or late eclampsia of the pregnant woman; and in combination with the prior probability and the conditional probability, calculating the posterior probability of premature or late eclampsia of the pregnant woman based on a Bayesian model. The detection of related protein markers and the acquisition of maternal clinical information data are relatively easy, the accuracy of predicting early-onset and late-onset preeclampsia is relatively high, clinical popularization and application are facilitated, and the kit has an important value for prevention and early diagnosis of diseases.

Description

A method for constructing a risk assessment and prediction model for preeclampsia and its application Technical Field

[0001] The present invention relates to the field of preeclampsia analysis, and specifically to a method for constructing a preeclampsia risk assessment prediction model, a prediction model, a prediction product, a prediction system and related applications thereof. Background Art

[0002] Preeclampsia (PE) and its concurrent hypertensive disorders are the second leading cause of maternal mortality, accounting for approximately 14% of all maternal deaths. PE is defined as new-onset hypertension with proteinuria or other organ damage (such as the kidney, liver, or brain) occurring after 20 weeks of gestation. Typical symptoms include hypertension, proteinuria, and edema. The global prevalence of PE is approximately 2% to 4%, with a current prevalence of approximately 2.3% in my country, but the proportion of severe preeclampsia is much higher than in other countries. Preeclampsia can be divided into two categories based on onset: early-onset preeclampsia (less than 34 weeks of gestation) and late-onset preeclampsia (≥34 weeks of gestation). Due to its heterogeneity and complexity, the primary treatment approach remains termination of pregnancy and delivery of the placenta. Risk prediction and early diagnosis are also challenging. PE can lead to serious complications, including internal bleeding, seizures, stroke, premature birth, and death, and can also lead to a range of postpartum complications for both the mother and the newborn. Currently, 60% of deaths from PE are due to missed or delayed diagnosis. Therefore, further research and development of earlier and more accurate clinical prediction methods for PE is of great significance for the early prevention and effective management of PE in pregnant women.

[0003] In recent years, many results have been published on the research on the prediction of preeclampsia using technical means including protein markers. "Method for determining the risk of prenatal complications" (CN102216468A) establishes a computer-implemented method for early prediction of the risk of preeclampsia in pregnant women. The invention determines the content of one or more biochemical markers in the serum of pregnant women, including placental growth factor (PLGF), pregnancy-related serum protein-A and placental protein 13 (Placental Protein 13, PP13) in one or more blood samples from individuals, performs median correction of protein concentration, measures the blood pressure of the individual, calculates the mean arterial pressure, measures the uterine artery pulsatility index of the pregnant woman, and uses the amount of each of the selected one or more biochemical markers and the individual's blood pressure, combined with the pregnant woman's own clinical information, and using a multivariate Gaussian distribution model to calculate and output the predicted risk of early-onset and late-onset eclampsia in pregnant women. "Dynamics of the ratio of soluble fms-like tyrosine kinase 1 (sFlt-1) or Endoglin / PLGF as an indicator of critical pre-eclampsia and / or HELLP syndrome" (CN103917875A) relates to a method for diagnosing whether a pregnant subject is at risk of developing pre-eclampsia in the short term, the method comprising: a) determining the amount of the biomarker sFlt-1 or Endoglin and the amount of PLGF in a first and a second sample of the subject, wherein the first sample is obtained before the second sample; b) calculating a first ratio from the amount of sFlt-1 or Endoglin and the amount of PLGF determined in the first sample, and calculating a second ratio from the amount of sFlt-1 or Endoglin and the amount of PLGF determined in the second sample; c) comparing the values ​​of the first ratio and the second ratio, wherein if the value of the second ratio increases by at least about 3 times compared to the value of the first ratio, the subject is diagnosed as having a risk of developing pre-eclampsia in the short term.

[0004] However, the above-mentioned method (CN102216468) needs to use the uterine artery pulsation index of pregnant women as an indicator for preeclampsia risk assessment. However, the determination of the uterine artery pulsation index requires high professionalism and requires the professional operation of an ultrasound physician, which is not conducive to the widespread application of the invention. The method (CN103917875) uses sFLT-1 or ENG / PLGF ratio to evaluate the risk of preeclampsia in pregnant women in the short term (less than 4 weeks) in the second and third trimesters, which cannot be used for the risk prediction of preeclampsia in early pregnancy, nor can it guide clinicians to use aspirin intervention in early pregnancy to reduce the risk of eclampsia. Therefore, it is of great significance to find more representative markers to achieve a more general, easy-to-implement and development of a prediction method.

[0005] Summary of the Invention

[0006] The present invention develops a new method for constructing a preeclampsia risk assessment and prediction model and its related applications through in-depth and systematic research on two preeclampsia-related protein markers, PLGF and IGFBP1, in the serum / plasma of pregnant women, and combining them with several clinical indicators and mean arterial pressure (MAP) of pregnant women.

[0007] In a first aspect, the present invention provides a method for constructing a preeclampsia risk assessment and prediction model based on several clinical indicators, MAP, and IGFBP1 and / or PLGF of pregnant women, including sick pregnant women with early-onset eclampsia or late-onset eclampsia and healthy pregnant women.

[0008] Specifically, the construction method includes:

[0009] Obtain the clinical indicators, MAP, and relevant data of IGFBP1 and / or PLGF of the pregnant woman as relevant modeling factors (that is, the above-mentioned modeling factors include clinical indicators and MAP, as well as at least one of IGFBP1 and PLGF), and construct a training sample set with these modeling factors and their corresponding pregnant women. The above-mentioned pregnant women include sick pregnant women with early-onset eclampsia and / or late-onset eclampsia and healthy pregnant women. Based on the above-mentioned training sample set, several types of model training are performed, and the several types of trained models are verified and evaluated, and the best prediction model is confirmed according to the model evaluation indicators. It should be understood that the above-mentioned prediction model can usually be evaluated by multiple indicators such as accuracy, precision, ROC curve, AUC, etc., and the best prediction model is confirmed based on the above-mentioned indicators.

[0010] In a specific embodiment of the present application, the above construction method includes (as shown in FIG8 ):

[0011] A logistic regression model was constructed based on the clinical indicators obtained from pregnant women to determine the prior probability of developing early-onset or late-onset eclampsia. Clinical indicators used in this logistic regression model included maternal age, BMI, parity, two or more miscarriages or induced labors, adverse medical history, in vitro fertilization, and multiple pregnancies.

[0012] The obtained MAP, IGFBP1, and / or PLGF values ​​of pregnant women were median-processed, and multivariate normal distribution models for diseased and healthy pregnant women were constructed based on the MAP MoM values, IGFBP1 MoM values, and / or PLGF MoM values, respectively, to obtain the conditional probability of pregnant women developing early-onset or late-onset eclampsia.

[0013] Combining the above prior probabilities and conditional probabilities, the posterior probability of early-onset or late-onset eclampsia in pregnant women was calculated based on the Bayesian model.

[0014] After obtaining the above prediction model, the ROC curve is constructed according to the obtained posterior probability, and the optimal cutoff value and the area under the curve are determined according to the ROC curve to determine the risk threshold.

[0015] In a second aspect, based on the first aspect, the present invention provides the use of protein markers IGFBP1 and / or PLGF in constructing a risk assessment and prediction model for preeclampsia.

[0016] In a third aspect, the present invention provides a kit for detecting protein markers associated with preeclampsia, wherein the protein markers include IGFBP1 and / or PLGF.

[0017] In a fourth aspect, based on the prediction model obtained in the first aspect, the present invention provides a method for assessing and predicting the risk of preeclampsia, the method comprising:

[0018] 1) Obtain clinical indicators, MAP, and IGFBP1 and / or PLGF data of the pregnant women to be tested;

[0019] 2) Inputting the obtained clinical indicators, MAP, and IGFBP1 and / or PLGF data of the above-mentioned pregnant women into the prediction model for processing to obtain the risk value of preeclampsia for the above-mentioned pregnant women. When the risk value is higher than the set threshold, the pregnant woman is determined to be at high risk of developing eclampsia.

[0020] In a specific embodiment of the present invention, for the prediction of premature eclampsia, the data of the clinical indicators of the above-mentioned pregnant women are input into a logistic regression analysis model to calculate the prior probability; the data of the above-mentioned MAP and IGFBP1 and / or PLGF are respectively input into the multivariate normal distribution model of pregnant women with premature eclampsia and healthy pregnant women to obtain the conditional probability; then the prior probability and conditional probability are input into the Bayesian model to calculate the posterior probability, and the premature eclampsia risk value of the pregnant woman to be tested is obtained.

[0021] In a specific embodiment of the present invention, for the prediction of late-onset eclampsia, the data of the clinical indicators of the above-mentioned pregnant women are input into a logistic regression analysis model to calculate the prior probability; the data of the above-mentioned MAP and PLGF are respectively input into the multivariate normal distribution model of pregnant women with late-onset eclampsia and healthy pregnant women to obtain the conditional probability; then the prior probability and conditional probability are input into the Bayesian model to calculate the posterior probability, and the risk value of late-onset eclampsia of the pregnant woman to be tested is obtained.

[0022] In a fifth aspect, based on the prediction method of the fourth aspect, the present invention provides a prediction system for preeclampsia risk assessment, the system comprising:

[0023] 1) a device for acquiring data related to clinical indicators, MAP, and IGFBP1 and / or PLGF of the pregnant woman to be tested;

[0024] 2) a device for performing predictive model processing on the clinical indicators, MAP, and IGFBP1 and / or PLGF data of the pregnant woman obtained in 1) above;

[0025] 3) A device for outputting prediction results.

[0026] In a sixth aspect, the present invention provides a product for predicting and assessing the risk of preeclampsia, comprising: a memory and a processor. The memory is configured to store a program; the processor is configured to execute the program stored in the memory to implement the method for predicting and assessing the risk of preeclampsia as described in the fifth aspect.

[0027] At the same time, the present invention also provides a computer-readable storage medium, on which a program is stored, and the program can be executed by a processor to implement the method for assessing and predicting the risk of preeclampsia as mentioned in the fourth aspect above.

[0028] In addition, the present invention also provides a computer-readable storage medium, which stores the prediction model obtained by the construction method of the first aspect.

[0029] The beneficial effects of the present invention are:

[0030] 1. The present invention establishes a predictive model for early pregnancy preeclampsia risk assessment based on protein markers and maternal clinical information. It does not rely on the uterine artery pulsatility index as an indicator, and achieves early pregnancy prediction of preeclampsia. The detection of relevant protein markers and the acquisition of maternal clinical information data are easier. The invention has a high accuracy in predicting early-onset and late-onset preeclampsia, is convenient for clinical promotion and application, is of great value for the prevention and early diagnosis of the disease, and is of great significance for improving maternal and infant outcomes and reducing maternal and infant mortality.

[0031] 2. The present invention constructs a risk model based on maternal clinical information and maternal MAP as well as PLGF and / or IGFBP1 to predict the risk assessment of preeclampsia in early pregnancy. A larger sample size (see the example) was used to evaluate the predictive effects of the above-mentioned early-onset preeclampsia and late-onset preeclampsia models. The results showed that the sensitivity and specificity of the early-onset preeclampsia prediction model were 84.62% and 90.02%, respectively, and the sensitivity and specificity of the late-onset preeclampsia prediction model were 52.94% and 90.02%, respectively. The model has good accuracy and has potential clinical application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 shows the comparison of MAP MoM values ​​among early-onset eclampsia, late-onset eclampsia, and healthy control samples in the training set;

[0033] Figure 2 shows the changes in protein concentration at different sampling days. Figure A shows that the expected median PLGF protein concentration increases significantly with increasing sampling day, and Figure B shows that the expected median IGFBP1 protein concentration decreases significantly with increasing sampling day.

[0034] Figure 3 compares the MoM values ​​of IGFBP1 and PLGF in the training set between eclamptic patients and healthy controls. Panel A shows that the median IGFBP1 protein concentration is significantly downregulated in both early-onset and late-onset eclampsia samples, and Panel B shows that the median PLGF protein concentration is significantly downregulated in both early-onset and late-onset eclampsia samples.

[0035] Figure 4 shows the prediction effect of the preeclampsia logistic prediction model in the training set;

[0036] Figure 5 shows the prediction effect of the preeclampsia prediction model in the training set;

[0037] Figure 6 shows the prediction effect of the preeclampsia logistic prediction model in the validation set;

[0038] Figure 7 shows the prediction effect of the preeclampsia prediction model in the validation set;

[0039] FIG8 is a schematic diagram of a method for constructing a risk assessment and prediction model for preeclampsia. DETAILED DESCRIPTION

[0040] Based on the limitations of existing technologies, the present invention further explores the protein markers and other more easily accessible maternal clinical indicators through the development of preeclampsia protein markers in order to obtain a more accurate prediction model for preeclampsia risk assessment. +6 The protein levels in peripheral blood serum / plasma samples from 2 weeks after surgery were analyzed, and two protein markers that can be used to predict the risk of preeclampsia were discovered and verified, including IGFBP1 and PLGF. The physiological significance of the two indicators is shown in Table 1.

[0041] Table 1 Detection indicators and clinical significance

[0042] The above-mentioned protein markers PLGF and IGFBP1 can be detected by various methods. In the embodiments of the present invention, the protein detection technology used is ELISA enzyme-linked immunosorbent assay technology. Those skilled in the art will understand that other protein detection technologies, such as protein arrays, proteomics, expression proteomics, mass spectrometry (e.g., liquid chromatography-mass spectrometry (LC-MS), multiple reaction monitoring (MRM), selected reaction monitoring (SRM), scheduled MRM, scheduled SRM), 2D PAGE, 3D PAGE, electrophoresis, proteome chips, proteome microarrays, Edman degradation, direct or indirect ELISA, immunosorbent assay, immunoPCR, proximity extension analysis, Luminex analysis or homogenization analysis, time-resolved fluorescence (TRF), fluorescent oxygen channeling immunoassay (FOCI) or luminescent oxygen channeling immunoassay, liquid mass spectrometry targeted protein detection technology, chemiluminescent immunoassay protein detection technology, fluorescent immunoassay protein detection technology may all produce similar detection results.

[0043] Furthermore, the present invention develops a new method for constructing a predictive model for the risk assessment of preeclampsia by incorporating relevant data on the protein markers PLGF and IGFBP1, mean arterial pressure (MAP, calculated by measuring blood pressure in both arms of an individual), and other maternal clinical indicators of pregnant women, including maternal age, body mass index (BMI), parity, two or more miscarriages or induced labors, in vitro fertilization, previous history of gestational diabetes, preeclampsia, chronic hypertension, systemic lupus erythematosus and antiphospholipid syndrome, and multiple pregnancies, as modeling factors. It should be understood that in the present invention, mean arterial pressure is of particular importance compared to general maternal clinical information; therefore, mean arterial pressure (MAP) is separately categorized in the present invention.

[0044] Based on the above-mentioned inventive concept, pregnant women were divided into two case groups (including early-onset and late-onset preeclampsia groups) and a healthy control group based on whether they had early-onset or late-onset preeclampsia. The control group consisted of randomly selected pregnant women of the same period, who were full-term pregnancies with no pregnancy complications, whose fetuses were growing well at birth, and who had no obstetric, medical, or surgical complications during pregnancy.

[0045] Obtain the modeling factors of the three groups of pregnant women. Specifically, after obtaining the raw data required for the above, perform certain data transformations. 1) Retrieve the clinical index information of the three groups of pregnant women, including the age, height, gestational age, +6Clinical data such as weight, gravidity, parity, adverse pregnancy history, adverse past history (including previous gestational diabetes, preeclampsia, chronic hypertension, systemic lupus erythematosus, and antiphospholipid syndrome), the presence of in vitro fertilization, and multiple pregnancies were subsequently used to construct a logistic regression model. The corresponding data transformation or assignment was specifically described in the subsequent logistic regression equation.

[0046] 2) The mean arterial pressure (MAP) of the three groups of pregnant women was tested, and the median multiple calibration of the mean arterial pressure values ​​of each pregnant woman was performed to obtain the MAP MoM value of each pregnant woman. Specifically, the median multiple calibration of the mean arterial pressure values ​​of the pregnant women was completed using the median of the healthy pregnant women in the control group. The MAP test includes: before the pregnant woman's blood is drawn, she sits and rests for 5 minutes, and then uses an electronic blood pressure monitor to measure the blood pressure of both arms at least twice. The systolic and diastolic blood pressures are measured to calculate the MAP value. The calculation formula (Formula 1) is as follows:

[0047] Among them, diastolic blood pressure (DBP) is diastolic blood pressure, and PP is systolic blood pressure minus diastolic blood pressure.

[0048] The MAP calibrated using the median MAP of the population is calculated using the following formula (Formula 2):

[0049] MAP is the original value of mean arterial pressure measured in pregnant women, P (Median) MAP is the median MAP of the population. In a specific embodiment of the present invention, the median MAP of the population is calculated using the data of healthy pregnant women as a control and assigned a value of 82.98. MAP is the median-corrected MAP MoM value.

[0050] 3) Measure the IGFBP1 and PLGF protein concentrations in serum / plasma samples from the three groups of pregnant women. Use the IGFBP1 and PLGF protein levels of healthy pregnant controls and the gestational day of sampling to construct linear equations and perform median multiple calibration of the original concentrations of the two protein indicators. The linear equation for the median expected IGFBP1 protein concentration at each gestational day (Formula 3) is as follows: E(Median) IGFBP1,GA =1.94×10 5 -1.19×10 3 ×GA Formula 3

[0051] The linear equation (Formula 4) for the median expected PLGF protein concentration at each gestational day is as follows: E(Median) PLGF,GA =-10.3+0.672×GA Formula 4

[0052] GA is the gestational age, E (Median) IGFBP1,GA and E(Median) PLGF,GA The median expected IGFBP1 and PLGF protein concentrations for each gestational day were calculated. The raw values ​​of IGFBP1 and PLGF protein concentrations for each pregnant woman were then divided by the median expected protein concentrations for the corresponding gestational day to calculate the MoM value (Formula 5).

[0053] Concentration protein is the original value of IGFBP1 and PLGF protein concentration for each pregnant woman, E (Median) protein, GA is the median of expected protein concentration on the corresponding gestational day, MoM protein MoM values ​​of protein median were corrected for gestational days.

[0054] Regarding the construction of the logistic regression model and the calculation of the prior probability, in a specific embodiment of the present invention, the modeling factor in 1) above, the clinical indicator information of the pregnant woman, is used to construct a logistic regression model to calculate the prior probability of early-onset and late-onset eclampsia in the pregnant woman. Specifically, the clinical indicator-related data in the aforementioned training sample set are used to construct the early-onset and late-onset eclampsia logistic regression equations (refer to Formula 6), and the independent variables used in this equation (i.e., X1 to X7 in the formula) include age, BMI, parity, whether there have been more than two miscarriages or induced labors, whether there is a bad past history (including previous gestational diabetes, history of preeclampsia, familial history of preeclampsia, history of chronic hypertension, systemic lupus erythematosus and antiphospholipid syndrome), whether there has been in vitro fertilization and whether there has been multiple pregnancies.

[0055] In this example, by substituting the relevant data in the training set into the above formula 6, the specific logistic regression equation applicable to this application is obtained as follows:

[0056] (1) The logit function (Logit_epe) of pregnant women with premature eclampsia is:

[0057] Logit_epe = -9.04 + twins * 19.06 + parity * -0.01 + IVF * 2.08 + more than two miscarriages or induced labors * 1.29 + past medical history * 2.63 + median adjusted age * 2.16 + median adjusted BMI * 4.32

[0058] Based on the above logit function, the prior probability of a pregnant woman developing premature eclampsia (Prior_prob_epe) is: Prior_prob_epe = exp(Prior_prob_epe) / (1+exp(Prior_prob_epe))

[0059] (2) The logit function (Logit_lpe) of pregnant women with late-onset eclampsia is:

[0060] Logit_lpe = -6.91 + parity * -0.19 + IVF * 1.13 + more than 2 miscarriages or induced labors * 1.05 + past medical history * 2.49 + median adjusted age * 1.2 + median adjusted BMI * 4.07

[0061] Based on the above logit function, the prior probability of a pregnant woman developing late-onset eclampsia (Prior_prob_lpe) is: Prior_prob_lpe = exp(Prior_prob_lpe) / (1+exp(Prior_prob_lpe))

[0062] In the calculation formula of the above logit function, the specific assignment of clinical indicator information of pregnant women is as follows:

[0063] The twin variable is 0 or 1, representing a single or twin pregnancy, respectively. Parity is the actual parity of the pregnant woman before the delivery of this pregnancy. The IVF variable is 0 or 1, representing whether the pregnant woman's current pregnancy was a natural pregnancy or IVF. The value of miscarriage or induced labor is 0 or 1, representing no miscarriage or induced labor, respectively. Past medical history is 0 or 1, representing no or a history of one or more of gestational diabetes, preeclampsia, family history of preeclampsia, chronic hypertension, systemic lupus erythematosus, and antiphospholipid syndrome. Age and BMI are age and BMI MoM values ​​calibrated to the median age and BMI of the population. The calculation formula (Formula 7) is as follows:

[0064] Among them, Clin is the maternal age or BMI original value, P (Median) Clin The median age or BMI of the population (calculated using the data of healthy pregnant women) was assigned 29 and 20.69, respectively. Clin MoM values ​​are median-corrected age or BMI.

[0065] Regarding the construction and calculation of conditional probabilities of multivariate normal distribution models for early-onset eclampsia, late-onset eclampsia, and healthy controls, the modeling factors in 2) and 3) above—MAP MoM, IGFBP1 MoM, and PLGF MoM—were used to construct multivariate normal distribution models for early-onset eclampsia, late-onset eclampsia, and healthy controls, respectively, to obtain the conditional probability of a pregnant woman developing early-onset or late-onset eclampsia. Specifically, the present invention uses the probability density function of the multivariate normal distribution to calculate the conditional probability of the sample, and its calculation equation (Formula 8) is as follows:

[0066] Among them, |∑| represents the determinant of the covariance matrix of X, ∑ -1 represents the inverse matrix of the covariance matrix of X, μ is the sample mean, and x represents the logarithm of the MoM value of the corresponding indicator (the mean parameters and covariance matrix parameters of the three multivariate normal distribution models are calculated using the data of the training sample set, and the relevant parameters specifically correspond to Example Tables 4-7).

[0067] It should be noted that, when constructing a multivariate normal distribution model, the present invention can make full use of MAP MoM, IGFBP1 MoM, and PLGF MoM values ​​and their corresponding various combinations to carry out specific model construction, and then screen the model based on the subsequent posterior probability related model evaluation results (for example, Example Tables 12-13). The preferred multivariate normal distribution model for early-onset eclampsia in the present invention is a three-dimensional model, that is, using three random variables, MAP MoM, IGFBP1 MoM, and PLGF MoM values, as inputs in the multivariate normal model, while the preferred multivariate normal distribution model for late-onset eclampsia is a two-dimensional model, using two random variables, MAP MoM value and PLGF MoM value, as inputs in the multivariate normal model.

[0068] Regarding the establishment of the Bayesian model and the calculation of the posterior probability, the above-mentioned prior probability and conditional probability are substituted into the Bayesian model to calculate the posterior probability of early-onset eclampsia or late-onset eclampsia in pregnant women and obtain a prediction model. The specific calculation formula (Formula 9) is shown below:

[0069] Where P(A=disease) is the prior probability of a pregnant woman developing early-onset eclampsia or late-onset eclampsia, calculated using a logistic retrospective model; P(A=normal) is 1-P(A=disease).

[0070] P(B=disease|A=disease) is the conditional probability of a pregnant woman with early-onset eclampsia or late-onset eclampsia calculated by the multivariate normal distribution model; P(B=disease|A=normal) is the conditional probability of a pregnant woman with early-onset eclampsia or late-onset eclampsia calculated by the multivariate normal distribution model.

[0071] P(A=disease|B=disease) is the posterior probability of a pregnant woman developing early-onset eclampsia or late-onset eclampsia calculated by the Bayesian model.

[0072] It should be noted that the above description mentions that the MAP MoM, IGFBP1 MoM, and PLGF MoM values ​​and their corresponding various combinations can be fully utilized in the multivariate normal distribution model to construct a specific model. Substituting these different models into the Bayesian formula can obtain multiple sets of different prediction models, as well as multiple sets of different ROC curves (Receiver Operating Characteristic Curve, ROC). Based on the different ROC curves, the sensitivity, specificity, positive predictive value, and negative predictive value of each model for predicting early-onset eclampsia and late-onset eclampsia can be calculated. The optimal model can be selected. At the same time, the ROC curve is constructed based on the posterior probability of a pregnant woman developing early-onset eclampsia or late-onset eclampsia calculated by the Bayesian model. The optimal cutoff value and area under the curve (AUC) are determined based on the ROC curve, and then the risk threshold is determined. This risk threshold is the standard for judging the risk value (posterior probability) output by the model when using the model to predict the risk of preeclampsia in pregnant women to be tested. Specifically, if the risk value exceeds the set risk threshold for early-onset eclampsia or late-onset eclampsia, the pregnant woman to be tested is judged to be at high risk of developing early-onset eclampsia or late-onset eclampsia.

[0073] It should be understood that after the model is constructed by the above method, when performing the risk assessment and prediction of preeclampsia, the clinical indicators, MAP, and IGFBP1 and / or PLGF data of the pregnant woman to be tested are input into the constructed logistic retrospective model, the multivariate normal distribution model of pregnant women with early-onset eclampsia or late-onset eclampsia, and the multivariate normal distribution model of healthy control pregnant women according to the processing method in the above model construction method, and the corresponding prior probability and conditional probability are calculated respectively. The obtained prior probability and conditional probability are then substituted into the Bayesian model to obtain the posterior probability of the pregnant woman. When the posterior probability of the pregnant woman to be tested exceeds the set high-risk threshold for early-onset eclampsia or late-onset eclampsia, the pregnant woman to be tested is determined to be at high risk of early-onset eclampsia or late-onset eclampsia. The above prediction method can effectively judge the risk of early-onset and / or late-onset preeclampsia in pregnant women during pregnancy, thereby realizing the prediction of the risk of preeclampsia in early pregnancy to guide clinical physicians to intervene in advance. In a specific embodiment of the present application, the threshold value of the early-onset eclampsia prediction model based on clinical indicator information + MAP + PlGF + IGFBP1 is 0.91, and the threshold value of the late-onset eclampsia prediction model based on clinical indicator information + MAP + PlGF is 0.88.

[0074] It should be understood that after the model is constructed by the above method, when performing preeclampsia risk assessment and prediction, the prediction models for early-onset eclampsia and late-onset eclampsia can be used independently, that is, the prediction of early-onset eclampsia and late-onset eclampsia can be performed independently for the same pregnant woman. At this time, the data input of the relevant pregnant women can be independent, that is, different protein data are input into the prediction models corresponding to different categories. Specifically, the data of the various clinical indicators, MAP, IGFBP1 and PLGF of the pregnant woman to be tested are input into the early-onset eclampsia model, and the data of the various clinical indicators, MAP and PLGF of the pregnant woman to be tested are input into the late-onset eclampsia model to predict the early-onset and late-onset risks of the pregnant woman. It should be understood that the two models can also be combined into one, and after uniformly inputting the two protein data of the pregnant woman as well as the various clinical indicators and MAP data, the prediction models are used to predict early-onset eclampsia and late-onset eclampsia respectively through data recognition and retrieval.

[0075] In addition, those skilled in the art should understand that in addition to the above-mentioned method of using the Bayesian formula to calculate the posterior probability to construct a prediction model, the above-mentioned clinical indicators of the pregnant woman, MAP, and IGFBP1 and / or PLGF data can also be used as input, and whether the pregnant woman suffers from preeclampsia as output, and the model can be constructed through machine learning, deep learning or artificial intelligence model substitution to obtain a preeclampsia risk assessment prediction model.

[0076] The present invention will be further described in detail below through specific embodiments with reference to the accompanying drawings.

[0077] Example:

[0078] 1. Sample Collection

[0079] 1.1 Inclusion criteria

[0080] Eligibility criteria for preeclampsia include: systolic blood pressure ≥140 mmHg and / or diastolic blood pressure ≥90 mmHg after 20 weeks of gestation, accompanied by any of the following: urine protein ≥0.3 g / 24 hours, urine protein / creatinine ratio ≥0.3, or spot urine protein ≥+ (test method when protein quantification is unavailable); absence of proteinuria but with involvement of any of the following organs or systems: heart, lungs, liver, kidneys, or abnormalities in the blood, digestive, or nervous systems, or placental-fetal involvement. Preeclampsia can be divided into two categories based on onset: early-onset eclampsia (premature eclampsia) before 34 weeks of gestation and late-onset eclampsia (late-onset eclampsia) ≥34 weeks of gestation.

[0081] Inclusion criteria for the healthy control group included uncomplicated full-term pregnancies, well-developed fetuses at birth, and no obstetric, medical, or surgical complications during pregnancy. Exclusion criteria included: ① other pregnancy complications; ② severe cardiac, hepatic, or renal insufficiency; ③ autoimmune diseases or malignant tumors. Pregnant women with abnormalities due to chromosomal or congenital anomalies, premature birth, or multiple pregnancies were also excluded.

[0082] 1.2 Participants:

[0083] The present invention adopts a retrospective case-control study method, retrieves the prenatal examination records and hospitalization history information of pregnant women during pregnancy through the clinical medical record system, and screens pregnant women according to the above-mentioned preeclampsia diagnostic criteria and healthy control inclusion criteria. A total of 113 cases of early-onset eclampsia, 198 cases of late-onset eclampsia and 2038 healthy control pregnant women were included in the study, as shown in Table 2. All pregnant women participated in non-invasive prenatal genetic testing, and the collected data from 11-15 gestational age were 48. +6 The remaining plasma samples were frozen at -80°C for subsequent biochemical analysis. The samples were divided into training and validation sets. The corresponding numbers of patients with early-onset eclampsia, late-onset eclampsia, and healthy controls for each dataset are shown in Table 2.

[0084] Table 2. Introduction to sample sources EPE: early-onset eclampsia, LPE: late-onset eclampsia, Control: healthy controls

[0085] 1.3 Clinical Information Analysis

[0086] Clinical information was compared between the training set of women with early-onset and late-onset eclampsia and healthy controls. Differences in continuous variables among the three groups were statistically analyzed using the Wilcoxon rank sum test, and differences in categorical variables were statistically analyzed using the Fisher exact test. The results showed that women with early-onset and late-onset eclampsia had significantly higher age, body mass index (BMI), and mean arterial pressure (MAP) than those in the control group (all P < 0.001). Women with early-onset and late-onset eclampsia also had a higher incidence of two or more miscarriages or induced labors, adverse medical history, IVF, and twin pregnancies (all P < 0.05). These results indicate that higher age, BMI, MAP, primiparas, a history of two or more miscarriages or induced labors, adverse medical history, IVF, and twin pregnancies are all high-risk factors for developing preeclampsia during pregnancy, significantly increasing the risk of both early-onset and late-onset preeclampsia (see Table 3).

[0087] Table 3. Comparison of clinical characteristics of pregnant women with early-onset and late-onset eclampsia and healthy controls in the training and validation sets *, **, and *** indicate P values ​​less than 0.05, 0.01, and 0.001, respectively. Adverse medical history included gestational diabetes mellitus, history of preeclampsia, family history of preeclampsia, history of chronic hypertension, systemic lupus erythematosus, and antiphospholipid syndrome.

[0088] 1.4 Mean arterial pressure measurement and calculation

[0089] After the pregnant woman rested in a sitting position for 5 minutes, her blood pressure was measured at least twice on both arms using an electronic blood pressure monitor (Omron J710). The measured systolic and diastolic blood pressures were input into the mean arterial pressure formula to calculate the MAP value.

[0090] Where diastolic blood pressure (DBP) is diastolic blood pressure, and PP is systolic blood pressure minus diastolic blood pressure. Each pregnant woman's mean arterial pressure (MAP) was then divided by the population median MAP of 82.89 (in this example, the population median MAP was calculated based on data from healthy pregnant controls) to generate the MAP MoM value. In the training set, MAP MoM values ​​for both early-onset and late-onset eclampsia samples were significantly higher than those for the healthy control group (P values ​​were all below 0.05, as shown in Figure 1).

[0091] 2. Median Multiple Calibration of Protein Marker Concentrations

[0092] The remaining plasma samples of the pregnant women were retrieved, and the concentrations of IGFBP1 protein and PLGF protein in the peripheral blood samples of each pregnant woman were determined by enzyme-linked immunosorbent assay. The present invention analyzed that the concentration of IGFBP1 protein in healthy pregnant women decreased significantly with the increase of sampling gestational age (correlation coefficient -0.22, P value <0.01), and the median concentration of PLGF protein increased significantly with the increase of sampling gestational age (correlation coefficient 0.18, P value <0.01), as shown in Figure 2. Based on the data of the above-mentioned healthy pregnant women, a linear equation of the expected median values ​​of IGFBP1 and PLGF proteins was constructed. The linear equation of the median expected IGFBP1 protein concentration on each gestational day is as follows, where GA is the number of gestational days: E (Median) IGFBP1,GA =1.94×10 5 -1.19×10 3 ×GA

[0093] The linear equation for the median expected PLGF protein concentration at each gestational day is as follows: E(Median) PLGF,GA =-10.3+0.672×GA

[0094] Each individual's IGFBP1 and PLGF protein concentrations were divided by the median expected protein concentration for the corresponding gestational day to perform median multiple calibration, resulting in the gestational day-calibrated protein median multiple MoM values. In the training set, the PLGF protein MoM value was significantly downregulated in both early-onset and late-onset eclampsia samples, and the IGFBP1 protein MoM value was significantly downregulated in both early-onset and late-onset eclampsia samples (all P values ​​were below 0.05, as shown in Figure 3).

[0095] 3. Calculate the prior probability of the training set samples

[0096] The age and BMI of each pregnant woman in the training set were divided by the group median age of 29 and the group median BMI of 20.69, respectively, to complete median correction. Based on the logistic regression equation in Formula 6 above, a specific logistic regression equation suitable for the present invention was constructed. The following specific logistic regression equation was used to calculate the prior probability of early-onset and late-onset eclampsia for each pregnant woman. The specific logistic regression equation suitable for the present invention obtained in this example is as follows:

[0097] Logit_epe = -9.04 + twins * 19.06 + parity * -0.01 + IVF * 2.08 + more than two miscarriages or induced labors * 1.29 + past medical history * 2.63 + median adjusted age * 2.16 + median adjusted BMI * 4.32

[0098] Logit_lpe = -6.91 + parity * -0.19 + IVF * 1.13 + 2 or more miscarriages or induced abortions * 1.05 + past medical history * 2.49 + median adjusted age * 1.2 + median adjusted BMI * 4.07 Prior_prob_epe = exp(Prior_prob_epe) / (1 + exp(Prior_prob_epe)) Prior_prob_lpe = exp(Prior_prob_lpe) / (1 + exp(Prior_prob_lpe))

[0099] In the above formula, Logit_epe and Logit_lpe represent the logit functions for the risk of early-onset and late-onset eclampsia, respectively. Prior_prob_epe and Prior_prob_lpe are the probabilities of early-onset and late-onset eclampsia, respectively, calculated using the logistic regression equations. The corresponding values ​​are described previously and are not repeated in this example.

[0100] At the same time, in order to better compare the model effects in the future, the AUCs of early-onset eclampsia and late-onset eclampsia in the training set samples were screened using the prior probability, and they were 0.78 and 0.75, respectively, as shown in Figure 4.

[0101] 4. Constructing a multivariate normal distribution model for pregnant women with early-onset and late-onset eclampsia and healthy controls

[0102] The mean arterial pressure of each pregnant woman in the training set was divided by the median mean arterial pressure of the population, 82.98 (same as 1.4), to complete median calibration. The median multiple calibration of PLGF and IGFBP1 was performed as described in 2 above. Then, the mean and covariance matrices of the products of mean arterial pressure MoM, PLGF MoM and IGFBP1 MoM, and PLGF MoM and IGFBP1 MoM of the multivariate normal distribution were established for pregnant women with early-onset eclampsia, late-onset eclampsia, and healthy controls, respectively. See Tables 4-7 below. The above data were substituted into the multivariate normal distribution model of pregnant women with early-onset eclampsia, late-onset eclampsia, and healthy controls (see Formula 8) to subsequently calculate the conditional probability of a pregnant woman developing early-onset or late-onset eclampsia.

[0103] Table 4. Mean values ​​of mean arterial pressure MoM, PLGF MoM, IGFBP1 MoM, and the product of PLGF MoM and IGFBP1 MoM in the multivariate normal distribution of pregnant women with early-onset eclampsia, late-onset eclampsia, and healthy controls

[0104] Table 5. Covariance matrix of mean arterial pressure MoM, PLGF MoM, IGFBP1 MoM, and product of PLGF MoM and IGFBP1 MoM in multivariate normal distribution of pregnant women with premature eclampsia

[0105] Table 6. Multivariate normal distribution of mean arterial pressure MoM, PLGF MoM, IGFBP1 MoM, and covariance matrix of the product of PLGF MoM and IGFBP1 MoM in pregnant women with late-onset eclampsia

[0106] Table 7. Covariance matrix of the product of mean arterial pressure MoM, PLGF MoM, IGFBP1 MoM, and PLGF MoM and IGFBP1 MoM in the multivariate normal distribution of healthy pregnant controls

[0107] Based on the variable combinations of the above multivariate normal distribution model, five prediction models are constructed using the same principle:

[0108] Model 1: clinical index information + MAP MoM value single link;

[0109] Model 2: clinical index information + MAP MoM value + IGFBP1 MoM value alone;

[0110] Model 3: clinical index information + MAP MoM value + PLGF MoM value alone;

[0111] Model 4: clinical index information + MAP MoM value + PlGF MoM value * IGFBP1 MoM value combination;

[0112] Model 5: clinical indicator information + MAP MoM value + PlGF MoM value + IGFBP1 MoM value triple.

[0113] 5. Using Bayesian models to calculate the posterior probability of early-onset and late-onset eclampsia

[0114] The prior probability of a pregnant woman developing early-onset or late-onset eclampsia calculated using the logistic retrospective model is combined with the conditional probability of a pregnant woman developing early-onset or late-onset eclampsia calculated using the multivariate normal distribution model for pregnant women with early-onset or late-onset eclampsia. The conditional probability of a pregnant woman developing early-onset or late-onset eclampsia is calculated using the multivariate normal distribution model for pregnant healthy controls. The posterior probabilities of early-onset and late-onset eclampsia are then calculated using the Bayesian model (see Formula 9). In this example, the posterior probabilities are calculated for each of the five prediction models mentioned in the above four.

[0115] 6. Evaluation of the accuracy of the present invention in screening early-onset and late-onset eclampsia in training set samples

[0116] Evaluation of the aforementioned prediction models 1-5 revealed that the triple early-onset eclampsia prediction model constructed by the present invention, combining clinical indicator information, MAP, PlGF, and IGFBP1, achieved an AUC of 0.86 for predicting the risk of early-onset eclampsia in the training set. When the threshold was set at 0.91, the corresponding sensitivity was 56% and the specificity was 95.58%, as shown in Table 8 and Figure 5. The clinical indicator information, MAP, and PlGF model achieved an AUC of 0.8 for predicting late-onset eclampsia. When the threshold was set at 0.88, the corresponding sensitivity was 39.04% and the specificity was 95.06%, as shown in Table 9 and Figure 5.

[0117] Table 8. Prediction results of premature eclampsia in training set samples

[0118] Table 9. Prediction results of late-onset eclampsia in training set samples

[0119] 7. Validation set sample validation Bayesian model

[0120] The validation set samples were used to further evaluate the prediction effect of the model constructed by the present invention on early-onset eclampsia and late-onset eclampsia. The results showed that the AUC of the early-onset eclampsia logistic prediction model for predicting early-onset eclampsia reached 0.74, and the AUC of the late-onset eclampsia logistic prediction model for predicting late-onset eclampsia reached 0.77, as shown in Figure 6.

[0121] The preeclampsia model established in the validation set also achieved high accuracy. The triple-combination model for predicting early-onset eclampsia, combining clinical markers, MAP, PlGF, and IGFBP1, had an AUC of 0.93. When the threshold was set at 0.91, the model's sensitivity and specificity were 84.62% and 90.02%, respectively. The positive and negative predictive values ​​for general screening were 6.25% and 99.87%, respectively (Table 10 and Figure 7). The model for predicting late-onset eclampsia, combining clinical markers, MAP, and PlGF, had an AUC of 0.83. When the threshold was set at 0.88, the model's sensitivity and specificity were 52.94% and 90.02%, respectively. The positive and negative predictive values ​​for general screening were 14.06% and 98.41%, respectively (Table 11 and Figure 7).

[0122] Table 10. Prediction results of premature eclampsia in validation set samples

[0123] Table 11. Prediction results of late-onset eclampsia in validation set samples

[0124] 8. Comparison of the effectiveness of different protein combination Bayesian models in screening for preeclampsia

[0125] Compare the accuracy of protein models for predicting early-onset eclampsia and late-onset eclampsia with different protein combinations (models 1-5) in the validation set.

[0126] The AUC, sensitivity, specificity, positive predictive value and negative predictive value of the triple model of maternal clinical indicator information + MAP + IGFBP1 + PLGF for screening premature eclampsia were better than those of the maternal clinical information alone and the maternal clinical information + single protein model.

[0127] In the late-onset eclampsia model, the combined maternal clinical indicator + MAP + IGFBP1*PLGF model performed better than the maternal clinical indicator information alone and the maternal clinical indicator information + MAP + IGFBP1 model, with the maternal clinical indicator information + MAP + PLGF model performing best. (See Table 12.)

[0128] These results show that the model constructed by the present invention performs better in predicting early-onset preeclampsia in early pregnancy.

[0129] It's important to understand that existing research data shows that the complication rate for women with early-onset eclampsia is much higher than that for those with late-onset eclampsia. In particular, because women with early-onset eclampsia often suffer liver and placental damage, the incidence of adverse perinatal outcomes is also significantly higher than that with late-onset eclampsia. Therefore, the significance and role of predicting early-onset eclampsia are even more important.

[0130] Table 12. Comparison of the effectiveness of different protein combinations in predicting early-onset and late-onset eclampsia in the validation set

[0131] The above examples are used to illustrate the present invention, which are only used to help understand the present invention and are not intended to limit the present invention. Those skilled in the art can make several simple deductions, modifications or substitutions based on the concept of the present invention.

Claims

1. A method for constructing a preeclampsia risk assessment and prediction model, characterized in that, the preeclampsia risk assessment and prediction model is constructed based on several clinical indicators, MAP, and IGFBP1 and / or PLGF of pregnant women, and the pregnant women include pregnant women with early-onset preeclampsia or late-onset preeclampsia and healthy pregnant women.

2. The construction method according to claim 1, characterized in that, the construction method includes: obtaining the clinical indicators, MAP, and IGFBP1 and / or PLGF of pregnant women as relevant modeling factors, and constructing a training sample set with the modeling factors and their corresponding pregnant women, and the pregnant women include pregnant women with early-onset preeclampsia or late-onset preeclampsia and healthy pregnant women; performing several types of model training based on the training sample set, evaluating the several types of trained models, and confirming the best prediction model according to the model evaluation index.

3. The construction method according to claim 2, characterized in that, the construction method includes: constructing a logistic regression analysis model based on the obtained clinical indicators of pregnant women to obtain the prior probability of pregnant women having early-onset preeclampsia or late-onset preeclampsia; the clinical indicators used in the logistic regression analysis model include the age, BMI, parity, having had more than 2 abortions or inductions, having a bad past history, having in vitro fertilization, and having multiple pregnancies of pregnant women; performing median processing on the obtained MAP, IGFBP1, and / or PLGF of pregnant women, and constructing a multivariate normal distribution model for pregnant women with preeclampsia and healthy pregnant women based on the MAP MoM value and the IGFBP1 MoM value and / or the PLGF MoM value respectively to obtain the conditional probability of pregnant women having early-onset or late-onset preeclampsia; combining the prior probability and the conditional probability, and calculating the posterior probability of pregnant women having early-onset or late-onset preeclampsia based on the Bayesian model.

4. The construction method according to claim 3, characterized in that, after obtaining the prediction model, constructing an ROC curve based on the posterior probability of the Bayesian model, and determining the best cut-off value and the area under the curve according to the ROC curve to determine the risk threshold.

5. The construction method according to claim 1, characterized in that, the detection method of the IGFBP1 and / or PLGF includes any one or a combination of ELISA enzyme-linked immunosorbent assay technology, targeted protein detection technology of liquid mass spectrometry, chemiluminescent immunoassay protein detection technology, and fluorescent immunoassay protein detection technology.

6. The construction method according to claim 1, characterized in that, The IGFBP1 and / or PLGF are from the peripheral blood serum / plasma of pregnant women at 11-15 +6 weeks.

7. The application of the protein markers IGFBP1 and / or PLGF in constructing a preeclampsia risk assessment and prediction model.

8. A protein marker detection kit related to preeclampsia, characterized in that, the protein markers IGFBP1 and / or PLGF.

9. A preeclampsia risk assessment and prediction method, characterized in that, the prediction method includes: obtaining the data of the clinical indicators, MAP, and IGFBP1 and / or PLGF of the pregnant woman to be tested; Input the data of the clinical indicators, MAP, and IGFBP1 and / or PLGF of the obtained pregnant woman into a prediction model for processing to obtain the preeclampsia risk value of the to-be-tested pregnant woman. If the risk value exceeds the threshold, it is determined that the to-be-tested pregnant woman has a high risk of eclampsia. The prediction model is obtained by the construction method described in any one of claims 1-6.

10. According to the prediction method described in claim 9, it is characterized in that for early-onset eclampsia, the threshold is 0.91; for late-onset eclampsia, the threshold is 0.

88.

11. According to the prediction method described in claim 9, it is characterized in that for early-onset eclampsia, input the data of the clinical indicators of the pregnant woman into a logistic regression analysis model to calculate the prior probability; input the data of MAP, IGFBP1, and PLGF into the multivariate normal distribution models of pregnant women with early-onset eclampsia and healthy pregnant women respectively to obtain the conditional probability; then input the prior probability and the conditional probability into a Bayesian model to calculate the posterior probability to obtain the early-onset eclampsia risk value of the to-be-tested pregnant woman; for late-onset eclampsia, input the data of the clinical indicators of the pregnant woman into a logistic regression analysis model to calculate the prior probability; input the MAP and the data of PLGF into the multivariate normal distribution models of pregnant women with late-onset eclampsia and healthy pregnant women respectively to obtain the conditional probability; then input the prior probability and the conditional probability into a Bayesian model to calculate the posterior probability to obtain the late-onset eclampsia risk value of the to-be-tested pregnant woman.

12. A prediction system for preeclampsia risk assessment, it is characterized in that the prediction system includes: a device for acquiring the data of the clinical indicators, MAP, and IGFBP1 and / or PLGF of the to-be-tested pregnant woman; a device for performing prediction model processing on the data of the clinical indicators, MAP, and IGFBP1 and / or PLGF of the pregnant woman; a device for outputting the prediction result; the prediction model is obtained by the construction method described in any one of claims 1-6.

13. A prediction product for preeclampsia risk assessment, it is characterized in that it includes: a memory for storing programs; a processor for implementing the method described in any one of claims 9-11 by executing the programs stored in the memory.

14. A computer-readable storage medium, it is characterized in that a program is stored on the medium, and the program can be executed by a processor to implement the method described in any one of claims 9-11.

15. A computer-readable storage medium, it is characterized in that a prediction model obtained by the construction method described in any one of claims 1-6 is stored on the medium.

Images