A Reappraisal of Second Trimester Mean Arterial Pressure as a Predictor of Pregnancy Induced Hypertension

  1. Homepage
  2. A Reappraisal of Second Trimester Mean Arterial Pressure as a Predictor of Pregnancy Induced Hypertension

Articles | Dr. Michael Rogers


The objective of our study was to evaluate the predictive ability of second trimester mean arterial pressure (MAP) for proteinuric and non-proteinuric pregnancy induced hypertension. A prospective study of MAP in pregnancy was performed on 220 normotensive, primagravid, Chinese women, recruited in the second trimester of pregnancy. MAP was measured in the left lateral position after rest using a Dinamap automatic sphygmomanometer. Patients were followed until delivery and classified as either having remained normotensive, or having developed pregnancy induced hypertension, with or without proteinuria. Thirty-four pregnancies later developed pregnancy hypertension, of which 15 had significant proteinuria, either antenatally (nine cases) or in labour (six cases). Receiver operating characteristic curve analysis confirmed earlier reports that second trimester MAP is a good predictor of proteinuric pregnancy hypertension, hut less able to predict non-proteinuric hypertension. Only one out of the 15 cases of proteinuric pregnancy hypertension had a second trimester MAP value below 68 mm Hg. We conclude that second trimester MAP predicts proteinuric better than nonproteinuric pregnancy-induced hypertension when measured under these conditions.


THE use of second trimester mean arterial pressure (MAP) as a predictor of pregnancy induced hypertension has been reported by several authors (Fallis and Langford, 1963; Page and Christianson, 1976a,b; Friedman and Neff, 1977; Quaas et al., 1982; Moutguin et al., 1985). Fallis and Langford first demonstrated that patients destined to develop preeclampsia tended to have higher average second trimester MAP than those who remained normotensive, despite blood pressure recordings in the normal range. They reported a sensitivity of 82 per cent and specificity of 88 per cent. Several subsequent reports have supported these findings but have failed to attain the same levels of sensitivity and specificity. Page and Christianson (1976b) prospectively studied 14 833 patients (both nulliparous and multiparous) and reported a sensitivity of 43 per cent and specificity of 87 per cent for MAP >/= 90 mm Hg. Friedman and Neff (1977) reporting on patients from the Collaborative Perinatal Project, noted a higher sensitivity and specificity for predicting `transient’ (non-proteinuric) than proteinuric pregnancy hypertension. Villar and Sibai (1989) noted that the sensitivity of MAP for predicting pre-eclampsia was only 8 per cent, and Dekker and Sibai (1991) in a recent review on prediction concluded that if a raised second trimester MAP predicted anything it is transient hypertension rather than pre-eclampsia.
The disparity in sensitivity and specificity between different reports may relate to differences in definition of pre-eclampsia, or merely reflect differences in the `cut-off’ used to discriminate between those at high and low risk. With the advent of preventive treatment in the forms of low dose aspirin (Wallenburg et al., 1986) and calcium supplements (Belizan et al., 1988) a real need has arisen for adequate predictive tests, so that treatment can be applied rationally. We therefore decided to re-evaluate second trimester MAP as a predictor of proteinuric and non-proteinuric hypertension in our predominantly Chinese population.

Materials & Methods

Two hundred and twenty normotensive, primigravid, Chinese women were recruited in the early second trimester. They were all given ultrasonic examination for confirmation of gestational age and exclusion of congenital malformations. The details of the proposed research were explained to them and consent obtained.
Patients were admitted (as day patients) in the second trimester between 18 and 26 weeks of pregnancy. On admission they were rested sitting for 30 minutes. Blood pressure was measured initially by conventional mercury column sphygmomanometry in the sitting position to exclude hypertension. The patient was then rested for a further 5 to 10 minutes in the left lateral recumbent position. Systolic (SBP), diastolic (DBP) and mean arterial blood pressure (MAPd) were measured in the right arm, which rested on the patient’s side, using a `Dinamap’ automatic sphygmomanometer (Dinamap 845A, Critikon Inc., Johnson and Johnson, Tampa, Florida) at I minute intervals until the blood pressure stabilised. The Dinamap was used for this study as the staff of our unit are familiar with this machine and we wished to eliminate any inter-observer variation associated with blood pressure measurement. After investigation the patients were allowed home and followed up in the normal antenatal clinic.

Mean arterial pressure was also computed (MAPc) from the systolic and diastolic pressures recorded from the Dinamap using the formula:

MAPc = DBP + (SBP – SBP / 3

Comparison of MAPd and MAPc was performed by paired t testing.
Patients were followed up until delivery. They were then divided into three groups: Group 1: those who remained normotensive (blood pressure > 140/90 mm Hg) throughout pregnancy and delivery; Group 2: those with non-proteinuric hypertension (blood pressure >/= 140 mm Hg systolic and >/= 90 mm Hg diastolic but proteinuria < 03 g/l); Group 3: those with ante-partum or intra-partum pregnancy induced hypertension with significant proteinuria (blood pressure >/= 140 mm Hg systolic and >/= 90 mm Hg diastolic with proteinuria >/= 0.3 g/l).
Blood pressure for the diagnostic categories was measured in the semi-recumbent supine position using conventional mercury column sphygmomanometry and Korotkof sound IV marking the diastolic pressure. Two readings of over 140/90 mm Hg taken 4 hours apart were considered diagnostic of pregnancy induced hypertension. Significant proteinuria was diagnosed if two separate specimens of urine tested with dipsticks gave readings of 2 + or more.
Receiver operating characteristic curves were generated and analysed using software designed in the Department of Obstetrics and Gynaecology of the Chinese University. The receiver operating characteristic curve of a test demonstrates the `trade off possible between increased detection of diseased individuals (sensitivity) against increased mixclassification of non-diseased individuals (false positive rate). It plots the test’s sensitivity against its false positive rate: the cutoff between test positive and test negative is varied across the entire range of test results. The curve is therefore a test characteristic which is independent of disease prevalence. The curve for a useless test follows the diagonal (line of indifference) and has an area beneath it of 0.5. The curve for a perfect test follows the left and upper boundaries and has an area of 10. The statistical comparison of receiver operating characteristic curves uses the Wilcoxon statistic as an estimate of the area beneath each curve as recommended by Hanley and McNeil (1983). The statistical significance of the curves was determined by comparing the area beneath each curve with that of a useless test, calculating a Z score and consulting the appropriate table for the probability of this value occurring by chance (Siegel, 1953). For further details on the use of receiver operating characteristic curves in evaluating test performance we recommend the excellent review by Beck and Shultz (1986).
Differences in MAP between normotensive and hypertensive patients were assessed using the non-parametric Mann-Whitney test (Seigel, 1953) from the Statistical Package for Social Sciences: a non-parametric statistical test was applied as insufficient data points were available at each gestational age to establish that variables were normally distributed. Comparisons were made separately of normotensive versus non-proteinuric and normotensive versus proteinuric pregnancy induced hypertension.


Out of the original 220 patients recruited none were hypertensive at the time of admission and only one developed mild hypertension within the next 10 weeks. Altogether 34 patients (15 per cent) ultimately developed pregnancy hypertension during the latter part of their antenatal course. Significant proteinuria was observed in nine cases during the antenatal period and a further six during labour (having been non-proteinuric antenatally).
The gestational age range at testing was from 18 to 26 weeks (mean 22.0 +/- 1.8; s.d.). with no differences between the three groups.
Patients who remained normotensive were lighter in weight than those who developed proteinuric hypertension (P < 0 05) but similar to those who only developed hypertension. There were no differences in maternal age or gestational age at admission.
There were no differences between MAPd (direct readings) and MAPc (computed) (mean difference -0.56., 95 per cent confidence interval: -137 to 0.25); therefore the direct readings (MAPd) are reported.
Second trimester MAP was higher in the group developing proteinuric hypertension than in normotensive patients (P < 0.001), but, there was no difference between the normotensive and nonproteinuric hypertension groups. These differences were similar regardless of whether the MAPd or MAPc was used in the statistical analysis. Results are summarised in the Table.
Receiver operating characteristic curves representing second trimester MAP prediction of pregnancy induced hypertension, with and without significant proteinuria, are shown in the Figure. Sensitivity and specificity values from previous reports on nulliparous women, summarised in Dekker and Sibai’s (1991) review, are also included in the Figure for comparison. Analysis of the curves shows that MAP discriminates much better between the normotensive and proteinuric pregnancy hypertension groups (W = 0 78; SEw = 0.04; P < 0.0001) than between the normotensive and non-proteinuric pregnancy hypertension groups (W = 0 62; SEW = 0.06; P < 0.05). Furthermore the curve for prediction of pregnancy hypertension associated with significant proteinuria stops abruptly halfway along the upper border of the graph rather than terminating in the upper right corner, as is more usual. This indicates that, at least in the population tested, sensitivity of 100 per cent can be achieved whilst still identifying about 50 per cent of patients who have no risk of developing pre-eclampsia if a cut-off value for MAP of 60 mm Hg is used. The optimum predictive performance shown on the receiver operating characteristic curve is at a sensitivity of 0.93 (missing one case out of fifteen) and a specificity of 0.61 representing a cut-off value for MAP of 68 mm Hg. This appearance is maintained if only cases who developed proteinuria antenatally are considered.



Our data show the reverse pattern of prediction to that found in Friedman and Neff’s (1977) series; gestational hypertension being poorly predicted whilst the prediction of proteinuric hypertension was excellent. The prevalence of pregnancy induced hypertension in the study group was higher than that normally quoted for our population (5 to 10 per cent).
One advantage of receiver operating characteristic curve analysis is that the results are independent of disease prevalence, being based on sensitivity and specificity rather than positive and negative predictive values which are dependent on prevalence. The curve for second trimester MAP prediction of proteinuric hypertension is of considerable interest. With the exception of Fallis and Langford’s (1963) series the sensitivity and specificity points summarised by Dekker and Sibai (1991) all fall close to the curve for prediction of proteinuric hypertension, suggesting that differences in definition are not the cause of the differences in sensitivity and specificity reported. It is more likely that variation in the cut-off points used is the main cause for these differences. For example, Villar and Sibai (1989) reported the lowest sensitivity of 8 per cent at a specificity of 93 per cent (Youden index = 1 per cent) using a cut-off value of > 90 mm Hg. Our data confirms that this value merely reflects their extremely high cut-off value which naturally results in a low sensitivity and high specificity. If a lower cut-off value had been used the Youden index (Youden, 1950) might have been considerably higher.
The optimum cut-off value almost certainly varies between populations as the distribution of MAP is dependent on race and nutrition as well as on gestation (Murnaghan, 1986). There is also a circadian rhythm with higher pressures recorded during the afternoon and evening and the lowest pressures between midnight and 0400 hours
(Murnaghan, 1986). All our observations were made during the same time period from 1000 to 1200 hours in a fairly homogenous group of ethnic Chinese, in a narrow gestational window. There was no difference between the gestational age range in the three groups of patients. The Dinamap 845A monitor is a semi-portable blood pressure monitor which relies on detecting oscillation of the arterial wall, caused by pulsatile blood flow through the vessel. Several authors have performed comparisons between the Dinamap and mercury column sphygmomanometers.

It is considered to be as reliable as direct intraarterial measurements in non-pregnant and postpartum women (Selig et al., 1989). Its use in pregnancy is less well established. Walker (1987) reported that in pregnancy there is an average chance of the Dinamap reading 8 mm Hg below conventional sphygmomanometry using Korotkov phase IV. Our measurements were made in the left lateral position and are therefore about 15 mm Hg lower than would be found in the sitting position. MAP tends to fall on resting, which is one of the major problems involved in using it for screening; in a busy antenatal clinic patients may not be able to rest for the pre-requisite time for blood pressure to stabilise, leading to a high false positive rate.
Potentially the most useful finding of this study is the observation that the receiver operating characteristic curve for prediction of proteinuric hypertension terminates abruptly on the upper X axis rather than at the upper right corner as is more usual. This means that a proportion of the population studied (at least half) can be identified as having no risk of developing proteinuric hypertension rather than just being a low risk group.

This is of importance in planning trials of preventive treatment as a group of patients cart be identified who do not require treatment, thus concentrating on the `at risk’ group (raising the prevalence in the test population). This was not the situation where non-proteinutic (gestational) hypertention was concerned. The point of intersection corresponds to a cut-off of 63 mm Hg in our population, using the Dinamap. This cut-off value should not be adopted by other centres without further testing. We have noted that in undernourished women from the Asian subcontinent, eclampsia may still occur in late pregnancy despite extremely low second trimester MAP. Our population is well nourished and affluent hut most pregnant women are not significantly under or overweight for their height.
Our conclusions are only valid for measurements made in the left lateral position with the Dinamap and may not be reproducible with other machines or other positions. The inclusion of the six patients who developed proteinuria during labour can be criticised on the basis that such patients really belong to a separate subgroup of pregnancy induced hypertension, with different perinatal implications. Our experience with such patients is that they almost invariably have a good perinatal outcome with birth weights within the normal range for term. They do exhibit a particularly malignant form of hypertension which requires prompt intensive treatment to avoid significant material morbidity. We therefore feel that their inclusion is valid. The curve for prediction of the more familiar antenatal disease alone is very similar to that shown in the Figure; with a slightly higher area beneath the curve of 0.82 but slightly higher SEw, of 0.05. MAP is a derived measurement which is more dependent on systolic pressure than on diastolic. Neither of these measurements performed as well as MAP at predicting pregnancy induced hypertension.


Download attached PDF below for more details:



Add your article content here.


Dr. Michael Rogers

Author Since:  May 3, 2018