Method, program, and device for predicting the onset of pregnancy-induced hypertension or preeclampsia

Abstract

To provide an onset prediction method for predicting an onset of hypertension based on data of systolic pressure and diastolic pressure measured by pregnant women at home, etc. on a daily basis, onset prediction program and onset prediction device.SOLUTION: An onset prediction method of pregnancy hypertension or preeclampsia includes the steps of: acquiring data of systolic pressure and diastolic pressure in a prescribed period; calculating a feature amount based on the data; and predicting an onset based on the feature amount. The onset prediction method includes an average of the systolic pressure and an average of the diastolic pressure in the feature amount as well as including an inclination when the systolic pressure is linearly regressed on gestational days and / or Pearson's correlation coefficient of the systolic pressure and diastolic pressure.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a method, a program, and an apparatus for predicting the onset of pregnancy-induced hypertension or preeclampsia. [Background technology]

[0002] Preeclampsia is a general term for high blood pressure in pregnant women and related diseases, and consists of four pathological conditions: gestational hypertension, preeclampsia, severe preeclampsia, and pregnancy complicated by hypertension. Of these, gestational hypertension and preeclampsia are characterized by the fact that hypertension does not occur before pregnancy, but occurs when blood pressure rises during pregnancy. This condition has a poor obstetric prognosis, and delays in transporting the mother often result in emergency cesarean section, so there is a clinical need to establish a method for predicting the onset of the condition in advance. Many inventions relating to the prediction of pregnancy-induced hypertension and preeclampsia have been devised to date (for example, Patent Documents 1 to 6).

[0003] Patent Document 6 discloses a method for predicting the risk of developing postpartum preeclampsia and the like. This method uses information on the levels of (1) sFlt-1 and PIGF, or (2) endoglin and PIGF in the subject sample as a solution, and it has been shown that the prediction of the onset of preeclampsia based on the ratio of sFlt-1 to PIGF in serum is highly accurate. A study conducted on Westerners (Zeisler H., et al., N.Engl.J.Med.,2016,doi:10.1056 / NEJMoa1414838) showed that the negative predictive value for predicting onset within one week was 99.3%, the sensitivity was 80.0%, and the specificity was 78.3%. In response to this, the inventors conducted a study targeting Asians (Bian X., et al., Hypertension, 2019, doi: 10.1161 / HYPERTENSIONAHA.119.12760) and demonstrated that predicting the onset of preeclampsia based on the ratio of sFlt-1 to PIGF is effective in Asians as well as in Westerners.

[0004] The prediction methods discussed in the above inventions and research all require the collection of blood or other samples from subjects. Therefore, this should be done at a medical institution, etc., and pregnant women cannot perform predictions on a daily basis at home. Therefore, there is a problem that depending on the timing of onset, it may not be possible to detect the condition in advance.

[0005] Meanwhile, as disclosed in Patent Document 7, the present inventors have invented an apparatus that enables accurate and rapid evaluation of the blood pressure of a pregnant woman, taking into consideration the normal range that varies depending on the course of pregnancy. When a pregnant woman uses this device to measure her blood pressure at home, the device displays the normal range according to the number of days of pregnancy along with the measured value. Therefore, pregnant women can routinely check whether their blood pressure is within the normal range at the time of measurement. However, this device does not have the function to predict whether a pregnant woman will develop high blood pressure in the future from the time of measurement. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] Japanese Patent Publication No. 2020-038412 [Patent Document 2] Japanese Patent Application Publication No. 2019-041754 [Patent Document 3] Special Publication No. 2018-513961 [Patent Document 4] Japanese Patent Application Laid-Open No. 2016-014685 [Patent Document 5] Special Publication No. 2017-512826 [Patent Document 6] Special Publication No. 2017-504029 [Patent Document 7] Japanese Patent Application Laid-Open No. 2013-165831 Summary of the Invention [Problem to be solved by the invention]

[0007] The present invention aims to provide an onset prediction method, an onset prediction program, and an onset prediction device for predicting the onset of hypertension based on data on systolic blood pressure and diastolic blood pressure measured daily by pregnant women at home, etc. [Means for solving the problem]

[0008] The onset prediction method of the present invention is a method for predicting the onset of gestational hypertension or preeclampsia, and includes the steps of acquiring data on systolic blood pressure and diastolic blood pressure over a predetermined period of time, calculating feature values ​​based on the data, and predicting onset based on the feature values, and is characterized in that the feature values ​​include the average systolic blood pressure and the average diastolic blood pressure, as well as "(1) the slope when systolic blood pressure is linearly regressed with respect to the number of days of pregnancy" and / or "(2) Pearson's correlation coefficient between systolic blood pressure and diastolic blood pressure." That is, in the present invention, one or both of the values ​​in (1) and (2) can be used as the feature amount as needed. In the present invention, the step of predicting onset may involve calculating a weighted sum based on the feature amounts and comparing it with a threshold, and the threshold may be adjusted depending on the required sensitivity and specificity.

[0009] It is preferable that the onset prediction method of the present invention further includes a step of removing the influence of the number of days of pregnancy and / or season from the acquired data and normalizing it, and that the feature amount is calculated based on the normalized data. A pregnant woman's blood pressure varies depending on the length of pregnancy and the season. Therefore, eliminating these influences increases the accuracy of onset prediction. The present invention may further include a step of interpolating data for days on which measurements were not taken before calculating the feature values, and / or a step of normalizing the feature values ​​by excluding the influence of the number of days of pregnancy and / or the season after calculating the feature values.

[0010] An onset prediction program for executing the onset prediction method and an onset prediction device including this onset prediction program are aspects of the present invention. [Effects of the Invention]

[0011] According to the present invention, pregnant women can routinely predict the onset of pregnancy-induced hypertension and preeclampsia at home, etc. [Brief explanation of the drawings]

[0012] [Figure 1] 1 is a flowchart illustrating an example of a method for predicting onset according to the present invention. [Figure 2] This shows the relationship between gestational age and blood pressure. [Figure 3] The relationship between the day of the year and blood pressure is shown. [Figure 4] Here is an example of linear interpolation of missing values: [Figure 5] The distribution of random noise is shown. [Figure 6] 1 shows the ROC curve for discriminating hypertension in pregnant women. [Figure 7] An example of time series data of blood pressure in a group of normotensive pregnant women is shown below. [Figure 8] An example of time series data of blood pressure in a group of pregnant women who developed hypertension is shown below. [Figure 9] The graph shows the progress of scores in the group of normotensive pregnant women and the group of pregnant women who developed hypertension. DETAILED DESCRIPTION OF THE INVENTION

[0013] FIG. 1 is a flowchart showing an example of the onset prediction method according to the present invention. Programs and devices for executing the onset prediction method of the present invention include built-in programs of blood pressure measurement devices, as well as application programs (hereinafter referred to as apps) for smartphones, tablet terminals, and computer software. The following explanation will be given using a smartphone app as an example, and "this app" refers to a program for executing the onset prediction method.

[0014] In step S11, data on systolic blood pressure (SBP) and diastolic blood pressure (DBP) for a predetermined period is obtained from the pregnant woman who is the subject. For example, a pregnant woman may input the systolic blood pressure and diastolic blood pressure (hereinafter simply referred to as blood pressure) values ​​measured at home using a blood pressure measuring device into this app on her smartphone. There are no particular limitations on the blood pressure measurement device, and for example, if a commercially available sphygmomanometer is used, pregnant women can measure their blood pressure on a daily basis at home. Furthermore, data may be transmitted from the blood pressure measurement device to the application via wireless or wired communication. The blood pressure measurement period for obtaining time-series blood pressure data is preferably about one to six weeks, and may include days on which no measurement was performed or days on which multiple measurements were performed. It is more preferable that the measurements be taken at approximately the same time, but this is not a limitation. In the following example, we will acquire blood pressure data for four weeks, including the most recent measurement date.

[0015] In step S12, the blood pressure data acquired in step S11 is normalized. As shown in Figure 2, which shows the relationship between the number of days of pregnancy and blood pressure, and Figure 3, which shows the relationship between the season (total number of days per year) and blood pressure, blood pressure is prone to fluctuate depending on the number of days of pregnancy and the season. Therefore, it is preferable to remove the influence of gestational age and season from the acquired blood pressure data, and more preferably, this normalization step includes standardization. As an example, the formulas (1) to (10) for excluding or standardizing the effects of gestational age and season are shown below.

number

[0016] The above formulas (1) to (8) were obtained from the results of research conducted by the inventors to date, and specifically, were derived using data from 12 to 20 weeks of pregnancy for 339 normotensive pregnant women. Here, normal blood pressure refers to a systolic blood pressure of less than 140 mmHg and a diastolic blood pressure of less than 90 mmHg, and the above 339 people excluded those whose blood pressure was not measured in the week prior to the delivery date. Equations (1) to (4) are calculation formulas for the systolic blood pressure of normotensive pregnant women, and equations (5) to (8) are calculation formulas for the diastolic blood pressure. Formula (1) represents the number of days of pregnancy (number of days of pregnancy when the first day of the last menstrual period in a pregnant woman is set to 0), formula (2) represents the total day of the year (the day of blood pressure measurement when January 1st is set to 1), and formula (3) represents the average systolic blood pressure when the number of days of pregnancy and the total day of the year are specified. Equation (4) represents the standard deviation of systolic blood pressure when the number of days of pregnancy and the total number of days of the year are specified, and is approximated by a linear function of the mean value. Equation (5) represents the average diastolic blood pressure when the number of days of pregnancy is specified, equation (6) represents the number of days in the year, equation (7) represents the number of days of pregnancy and the number of days in the year, and equation (8) represents the standard deviation of diastolic blood pressure when the number of days of pregnancy and the number of days in the year are specified. On the other hand, equations (9) and (10) are calculation formulas for standardizing the blood pressure data acquired in step S11 by excluding the influence of the number of days of pregnancy and the season (total number of days per year) calculated using equations (1) to (8) above. The normalized systolic blood pressure (normalized SBP) is calculated using equation (9), and the normalized diastolic blood pressure (normalized DBP) is calculated using equation (10).

[0017] In step S13, missing values ​​are interpolated to interpolate data for days on which measurements have not been taken. As shown in Figure 4, missing values ​​can be interpolated by linearly interpolating (obtaining the values ​​shown by the white circles in the figure) the values ​​before and after the non-measurement days (values ​​shown by the black circles in the figure), and then adding random noise that follows a two-dimensional normal distribution, as shown in Figure 5. The noise covariance matrix S in this embodiment is as shown in the following formula (11), which was also obtained from the inventors' research results to date. The interpolated data is measured once a day.

number

[0018] In step S14, the feature amount is calculated. The feature quantities include the mean systolic blood pressure, the mean diastolic blood pressure, the slope of the linear regression of systolic blood pressure with respect to the number of days of pregnancy, and the Pearson correlation coefficient between systolic blood pressure and diastolic blood pressure. For example, the average systolic blood pressure refers to the average systolic blood pressure over the four weeks including the most recent measurement date. If the above four features, for example, indicate that the systolic blood pressure and diastolic blood pressure are high on average, the systolic blood pressure gradually increases in proportion to the number of days of pregnancy, and the movement of the diastolic blood pressure is similar to that of the systolic blood pressure, there is a high risk of developing hypertension in the near future. As calculation formulas for calculating the four feature amounts, the following formulas (12) to (18) are given using the normalized systolic blood pressure and the normalized diastolic blood pressure obtained in steps S12 and S13. The normalized systolic blood pressure and normalized diastolic blood pressure can be replaced with the values ​​of systolic blood pressure and diastolic blood pressure before normalization, respectively, but it is preferable to use the normalized values ​​from the viewpoint of the accuracy of onset prediction.

number

[0019] In step S15, the feature values ​​calculated in step S14 may be normalized by excluding the influence of the number of days of pregnancy and the season. As an example, the following formulas (19) to (30) are used to exclude or standardize the influence of the number of days of pregnancy from the four feature quantities. Equations (19) to (26) were also derived from the results of previous research, and it is believed that the feature values ​​are more susceptible to the influence of the number of days of pregnancy than the season.

number

[0020] In step S16, the onset of pregnancy-induced hypertension or preeclampsia is predicted. The onset prediction is performed by calculating a weighted sum based on the feature amount, and if the resulting score exceeds a threshold, it can be determined that there is a risk of onset after blood pressure measurement. For example, by presenting the results of onset predictions to pregnant women using this app, pregnant women who are predicted to develop high blood pressure can see a doctor at the appropriate time. As an example, the formula (31) for calculating the weighted sum of the four normalized feature amounts calculated in step S15 is shown below, and the parameters used were those calculated from the results of previous research. In this embodiment, the weighted sum of four feature amounts is calculated, but the parameters may be changed or the number of feature amounts may be reduced.

number

[0021] The threshold used for the determination may be adjusted depending on the required sensitivity and specificity. For example, Figure 6 shows the receiver operating characteristic curve (ROC curve) for 378 normotensive pregnant women and 65 hypertensive pregnant women (443 in total). From Figure 6, the false positive rate (FPR) and true positive rate (TPR) which are equal to 1 minus specificity can be read, and thresholds that result in the desired sensitivity and specificity can be selected. Table 1 shows typical values ​​such as threshold, FPR, and TPR as examples. Note that PPV (Positive Predictive Value) represents the positive predictive value, and NPV (Negative Predictive Value) represents the negative predictive value. [Table 1]

[0022] Figure 7 shows an example of time series data of blood pressure for a group of normal blood pressure pregnant women (Normal), and Figure 8 shows an example of time series data of blood pressure for a group of hypertensive disorders of pregnancy (HDP). The progress of the scores for both groups in this case is shown in FIG. The scores were calculated using the above formulas (1) to (31). FIG. 9 suggests that the score determined in this example is effective in distinguishing between normotensive pregnant women and those who develop hypertension. Furthermore, when the threshold was set to achieve a specificity of 90%, it was 3.95, with a negative predictive value of 96%, a positive predictive value of 54%, and a sensitivity of 77%. In this embodiment, the score was calculated using four features (average systolic blood pressure, average diastolic blood pressure, slope of linear regression of systolic blood pressure with respect to gestational age, and Pearson's correlation coefficient between systolic blood pressure and diastolic blood pressure), but the number of features may be reduced depending on the detection accuracy and settings. For example, if the Pearson correlation coefficient between systolic and diastolic blood pressure is omitted, it may be possible to omit the step of interpolating missing values.

[0023] The onset prediction results may be transmitted to the medical institution in charge via the Internet or the like from a program for executing the onset prediction method of the present invention, for example, a smartphone app. Furthermore, in the case of a blood pressure monitor that incorporates the onset prediction method of the present invention as a program, it is possible to perform blood pressure measurement, data storage, onset prediction calculation, and presentation of prediction results in the same device.

Claims

1. A method for predicting the onset of pregnancy-induced hypertension or preeclampsia, comprising: The method includes a step of acquiring data of systolic blood pressure and diastolic blood pressure for a predetermined period by the device, a step of calculating a feature amount based on the data by the device, and a step of predicting an onset of the disease based on the feature amount by the device, The feature quantity includes the average systolic blood pressure and the average diastolic blood pressure, as well as the slope of a linear regression of the systolic blood pressure with respect to the number of days of pregnancy and / or the Pearson correlation coefficient between the systolic blood pressure and the diastolic blood pressure, The onset prediction method is characterized in that the step of predicting onset comprises calculating a weighted sum based on the feature amount and comparing the sum with a threshold value to make a judgment.

2. The method for predicting onset of pregnancy described in claim 1, further characterized in that the device has a step of excluding and normalizing the effects of the number of days of pregnancy and / or the season from the data acquired, and the feature is calculated based on the normalized data, and / or the device has a step of interpolating data for days not measured before calculating the feature, and / or the device has a step of excluding and normalizing the effects of the number of days of pregnancy and / or the season after calculating the feature.

3. An onset prediction program for executing the onset prediction method according to claim 1 or 2.

4. An onset prediction device comprising the onset prediction program according to claim 3.

Images