Magnetic resonance spectroscopy in pre-eclampsia: evidence of cerebral ischaemia

Authors


*Correspondence: Professor P. Rubin, Dean's Office, Queen's Medical Centre, Nottingham NG7 2UH, UK.

Abstract

Objective To establish whether there are changes in the maternal brain in pre-eclampsia detectable by magnetic resonance angiography and spectroscopy.

Design A prospective, observational study.

Setting Obstetric and Radiology Departments, Queen's Medical Centre, Nottingham.

Sample Fourteen healthy, nulliparous non-pregnant women, 9 healthy primiparous pregnant women and 10 women with pre-eclampsia.

Methods Magnetic resonance angiography and proton magnetic resonance spectroscopy of the brain was performed on each woman. Non-pregnant women were each studied twice. Healthy pregnant women were studied three times during pregnancy and once postnatally. Subjects with pre-eclampsia were studied once antenatally and twice postnatally. Magnetic resonance angiograms were examined for signs of vessel narrowing. On magnetic resonance spectroscopy, the ratios of the dominant peaks of the spectrum: N-acetyl aspartate (NAA), choline, creatine and lactate were compared.

Main outcome measures Comparison of spectroscopic indices in non-pregnant, normal pregnant and pre-eclamptic women.

Results On magnetic resonance angiography, there was no evidence of vessel narrowing in any of the three groups. NAA/choline ratio was higher at all stages of pregnancy compared with the non-pregnant group (P < 0.05) associated with lower choline. NAA/choline increased gradually during healthy pregnancy associated with a decrease in choline. NAA/choline was significantly lower in the pre-eclampsia group compared with the healthy pregnant women at similar gestation (P < 0.01), associated with higher choline. There were no differences between the groups postnatally. Lactate was not detected. These changes are similar to those found in patients with carotid stenosis without cerebral infarction.

Conclusions Narrowing of vessels detectable on magnetic resonance angiography does not occur commonly in pre-eclampsia. Magnetic resonance spectroscopy results suggest that there is relative cerebral ischaemia in pre-eclampsia compared with healthy pregnancy.

Introduction

Pre-eclampsia is associated with a decrease in circulating plasma volume, evidence of generalised vasoconstriction, increased systemic vascular resistance and reduced end-organ perfusion1. Pathological studies of the brain in women who have died from eclampsia have suggested that there is cerebral ischaemia caused by cerebral vasospasm2. It has therefore been proposed that the seizures of eclampsia occur because of cerebral ischaemia as a result of reduced cerebral perfusion.

Reduced perfusion of tissues leads to disturbances in cellular metabolism that ultimately result in cell death. The reduction in oxygen supply to cells causes them to produce energy by anaerobic methods. Lactate, produced as a by-product of anaerobic metabolism, accumulates in areas of deficient circulation.

Magnetic resonance techniques are non-invasive methods of investigating the brain and are considered safe in pregnancy. Previous reports where magnetic resonance angiography has been performed in pre-eclampsia and eclampsia show conflicting results. Some report visible cerebral vasospasm3–6, whereas others show no differences in the diameter of the cerebral vessels in pre-eclampsia and eclampsia7.

Proton magnetic resonance spectroscopy is a non-invasive method of investigating cerebral metabolism and is suitable for use during pregnancy. Proton magnetic resonance spectroscopy detects compounds in the brain that can give an indication of altered metabolism. The main compounds detected by proton magnetic resonance spectroscopy are N-acetyl aspartate (NAA), choline-containing compounds, creatine plus phosphocreatine and lactate. Changes in these compounds have been detected in conditions where there is known cerebral ischaemia or infarction such as stroke and carotid artery stenosis8–12. Little is known of the physiological changes in cerebral metabolism as part of the normal adaptation to pregnancy.

We hypothesised that there may be visible cerebral vasospasm in pre-eclampsia on magnetic resonance angiography and that there may be evidence of disturbance of cerebral metabolism in pre-eclampsia, detectable by proton magnetic resonance spectroscopy, as a precursor of the cerebral ischaemia, which results in eclampsia.

We set out to determine whether there is detectable narrowing of the cerebral arteries in women with pre-eclampsia on magnetic resonance angiography and to compare the findings with those from non-pregnant women and women in healthy pregnancy. In addition, we wanted to identify whether there are any changes in the concentrations of the cerebral metabolites NAA, choline, creatine and lactate measured using proton magnetic resonance spectroscopy during the course of normal healthy pregnancy. We compared the concentrations of these cerebral metabolites in healthy pregnancy and pre-eclampsia.

Methods

Three groups of women were recruited: non-pregnant nulliparous women, healthy pregnant primiparous women and women with pre-eclampsia.

Fourteen nulliparous women were recruited (age range 18–37, mean 28). All were non-smokers. One subject was taking beclomethasone and salbutamol inhalers for asthma, one had a past history of lumbar disc prolapse and one had a previous lumbar sympathectomy for Raynaud's syndrome. Three women were currently taking the combined oral contraceptive pill and one was taking minocycline for acne. All other subjects had no known history of cardiovascular, neurological, renal or metabolic disease.

Nine healthy pregnant volunteers were recruited from the antenatal clinics at the University Hospital (age range 23–34, mean 30). All were less than 18 weeks of gestation at the time of the first study. All were non-smokers and none had a previous pregnancy that had progressed beyond 12 weeks of gestation. None of the participants was known to suffer from hypertension, cardiovascular, neurological or metabolic disease. Two women had mild asthma requiring the use of beclomethasone and terbutaline inhalers. None of the other subjects were taking long term medication.

Ten women with pre-eclampsia were recruited from the antenatal wards and labour suite (age range 20–38, mean 31). Pre-eclampsia was defined as blood pressure ≥ 140/90 mmHg on at least two occasions four hours apart or blood pressure >160/110 mmHg on a single occasion plus an increase from the first trimester of >30 mmHg systolic and 15 mmHg diastolic blood pressure together with proteinuria >0.3 g/24 hours or more than ‘++’ on dipstick testing.

One woman was multiparous and the remainder nulliparous. All were non-smokers at the time of recruitment but three had smoked until they discovered that they were pregnant. Two women had twin pregnancies and two had conceived as a result of in vitro fertilisation treatment. One woman had asthma and hypothyroidism and was taking salbutamol and beclomethasone inhalers and oral thyroxine. None of the other subjects was known to suffer from hypertension, cardiovascular, neurological or metabolic disease. None of the women was receiving magnesium sulphate at the time of study.

One woman with eclampsia was also recruited and had magnetic resonance angiography performed but was not included in the magnetic resonance spectroscopy study.

Ethical approval was obtained from the University of Nottingham Medical School Ethics Committee for the non-pregnant volunteers and from the University Hospital Ethics Committee for the studies on pregnant women. All subjects gave informed, written consent having received detailed verbal descriptions of the procedures and a written information sheet approved by the ethics committees.

Non-pregnant women had spectroscopy performed on two occasions at least nine weeks apart.

Healthy pregnant women were studied as follows:

Second trimester14–17 weeks of gestation
Early third trimester23–26 weeks of gestation
Late third trimester34–36 weeks of gestation
Late postpartum14–23 weeks postpartum

One woman withdrew after the first visit and one woman who fainted after her second study withdrew following that incident. All the other women completed four visits except for one who delivered at 37 weeks of gestation on the day before her scheduled late third trimester visit.

Women with pre-eclampsia were studied as follows:

Third trimesterantenatally on recruitment to the study with confirmed pre-eclampsia
Early postpartum2–6 days postdelivery
Late postpartum14–24 weeks postpartum

Magnetic resonance studies were carried out on a 1.5 T Magnetom Vision MR scanner (Siemens).

Magnetic resonance angiography was performed using a three-dimensional ‘time of flight’ technique (TR 39 ms, TE 6.5 ms, flip angle 20°, slice thickness 1.09 mm, matrix 168 × 256, field of vision 175 × 200°). Postprocessing was carried out using a maximum intensity projection procedure rotating around a left–right axis to produce 12 images of the circle of Willis over a 360° rotation.

For proton magnetic resonance spectroscopy, a chemical shift imaging hybrid technique was used. The volume of interest was positioned posteriorly in the midline of the brain. Care was taken to ensure that this area did not contain areas of subcutaneous fat. Interactive shimming was performed to optimise the homogeneity of the magnetic field to improve spectral resolution and increase the signal/noise ratio. If possible, the full width at half maximum was reduced to <6 Hz. To improve water suppression, transmitter amplitude was adjusted until the ADC-V (analogue digital converter voltage) value was <1 or as low as possible. The time taken to set up the spectroscopy was between 10 and 30 minutes. The acquisition time was 6 minutes and 30 seconds. Raw data and base images were stored on optical disk for later postprocessing.

Following transfer of the data to a UNIX workstation, Fourier transformation was performed and automatic postprocessing was carried out using a preprogrammed protocol that allowed for phase, baseline and frequency shift correction. Peaks were assigned to NAA, choline, creatine and lactate. Integral values and peak amplitudes were calculated for each metabolite in each voxel within the volume of interest. By looking at the position of the voxels within the volume of interest on the image, voxels were chosen which contained only white matter, did not overlap with the ventricles and were not in the border zone around the outside of the volume of interest where the spectra were less satisfactory because of chemical shift artefact. Usually, there were two such voxels on each side in the parietal area and two to four voxels on each side in the occipital area. The mean values of the integrals of NAA, choline and creatine for the selected voxels was calculated and used in the statistical analysis that was carried out using SPSS for Windows (SPSS, Chicago, IL, USA). Ratios of NAA/choline and NAA/creatine were calculated for each individual study using the mean values.

Median and interquartile ranges were calculated for the mean integral values of NAA, choline and creatine and for the ratios of NAA/choline and NAA/creatine for each group at each visit.

Box plots have been used to illustrate the magnetic resonance spectroscopy results. The box represents the 25th to 75th centile and the line within the box is the median. The whiskers represent the extent of the range of values excluding outliers and extremes. The Mann–Whitney U test was used to determine the statistical significance of differences between the groups and the Wilcoxon signed rank test was used to assess the statistical significance of the changes within groups.

Results

Data were not available on some subjects for some visits either because the woman could not tolerate being in the scanner long enough to be able to complete the scan or because there was a technical problem with the storage of the spectroscopy raw data onto optical disk. Data were also excluded from two visits in the pre-eclampsia group where the woman was taking antihypertensive medication. There was no significant difference in the ages of the women in the three groups. The gestation of delivery in the pre-eclampsia group was significantly earlier than the healthy pregnancy group.

None of the subjects from any of the three study groups had demonstrable narrowing of the cerebral vessels on magnetic resonance angiography. However, the one woman with eclampsia had visible narrowing of the right middle cerebral artery on her initial visit, which had resolved by five days postpartum (Fig. 1).

Figure 1.

Magnetic resonance angiograms from subject with eclampsia. (a) appearances <24 hours postdelivery with narrowing of middle cerebral artery; (b) five days postpartum showing resolution of vascular narrowing.

The results for proton magnetic resonance spectroscopy are shown in Figs 2 and 3. No significant lactate peaks were seen in any of the subjects studied.

Figure 2.

Box and whisker plots showing magnetic resonance spectroscopy metabolite ratios for (a) NAA/choline, (b) NAA/creatine and (c) choline/creatine non-pregnant, healthy pregnant and pre-eclampsia groups. Heavy horizontal line = median; box = interquartile range; whiskers = extent of range of values excluding outliers and extremes. inline image= Non-pregnant group; inline image= 14–17 weeks pregnant; inline image= 23–26 weeks pregnant; inline image healthy pregnancy 34–36 weeks, pre-eclampsia visit 1 (29–38 weeks); inline image= 14–26 weeks postpartum.

Figure 3.

Box and whisker plots showing magnetic resonance spectroscopy metabolite integrals for (a) NAA, (b) choline and (c) creatine non-pregnant, healthy pregnant and pre-eclampsia groups. Heavy horizontal line = median; box = interquartile range; whiskers = extent of range of values excluding outliers and extremes. inline image= Non-pregnant group; inline image= 14–17 weeks pregnant; inline image= 23–26 weeks pregnant; inline image= healthy pregnancy 34–36 weeks, pre-eclampsia visit 1 (29–38 weeks); inline image= 14–26 weeks postpartum.

In the non-pregnant group, there were no statistically significant differences in any of the parameters between visits 1 and 2 using the Wilcoxon signed rank test. For the comparisons with other groups, only the results for visit 1 have been used.

The NAA/choline ratio was significantly higher in the healthy pregnancy group for all the visits during pregnancy compared with the non-pregnant group (P < 0.05). This was highly significant for the second trimester and late third trimester visits (P < 0.01). There was no significant difference in the NAA/choline ratio between the healthy pregnant group at the late postpartum visit and the non-pregnant group. The differences in the NAA/choline ratio were accompanied by a significantly lower choline for the healthy pregnancy group at the second trimester and late third trimester visits compared with the non-pregnant group (P < 0.05). The choline at the early third trimester visit was also lower but the result did not reach statistical significance.

The NAA/creatine ratio for the healthy pregnancy group at both third trimester visits was significantly higher than the non-pregnant group (P < 0.05).

There were no other significant differences between the two groups.

There was a gradual increase in the NAA/choline ratio through normal pregnancy and a marked decrease at the postpartum visit. There were statistically significant differences between late third trimester and late postpartum and between second trimester and late postpartum (P < 0.05). This was associated with a significant reduction in choline at the late third trimester visit to the early third trimester visit and this was also significantly lower than the level of choline at the late postpartum visit (P < 0.05). There was also a smaller increase in the NAA/creatine ratio between the two third trimester visits and a corresponding reduction in this ratio at the postnatal visit. There were no other statistically significant differences between the groups.

Within the pre-eclampsia group there were no significant differences in the spectroscopy parameters between the third trimester and early postpartum visits. There was a significant increase in the choline/creatine ratio between the third trimester and late postpartum visits (P < 0.05 Wilcoxon signed rank test). This was accompanied by a significantly higher choline at the late postpartum visit compared with the third trimester visit (P < 0.05).

Comparisons were made between healthy pregnancy late third trimester visit (34–36 weeks: mean 35.5 weeks) and pre-eclampsia third trimester visit (29–38 weeks: mean 33.7 weeks) and also between the late postpartum visits in the healthy pregnancy (14–23 weeks postpartum) and pre-eclampsia (14–26 weeks postpartum) groups. There were no significant differences in the gestations of the two groups at these visits.

Antenatally in the third trimester, the NAA/choline ratio and NAA/creatine ratio were significantly lower in pre-eclampsia than in the healthy pregnant group (P < 0.01 and P < 0.05) and choline was significantly higher (P < 0.05). Creatine was also higher but this failed to reach statistical significance. There were no significant differences between the two groups at the late postpartum visit.

Discussion

Most of the women with pre-eclampsia in this study had mild or moderate disease. None of them had narrowing of the cerebral vessels on magnetic resonance angiography. The one woman with eclampsia had obvious narrowing of the right middle cerebral artery on scan within 24 hours of delivery that had resolved by five days postpartum. Review of the literature regarding visible vasoconstriction of the cerebral vessels in pre-eclampsia and eclampsia using traditional or magnetic resonance angiography suggests that, in keeping with these findings, vasospasm of the larger cerebral vessels is apparent in severe pre-eclampsia and eclampsia but not in mild or moderate disease.

The published reports of traditional angiography in pre-eclampsia and eclampsia demonstrating cerebral vessel narrowing13–15 include data from four patients, all of whom had severe neurological features such as coma or paraplegia. The early case reports of magnetic resonance angiography in pre-eclampsia and pre-eclampsia3,4 involved one woman with typical eclampsia and one patient with severe pre-eclampsia complicated by a right hemiparesis. In both of these women, there was reversible diffuse vasospasm of the intracerebral arteries.

In a study in which magnetic resonance angiography was performed on 10 women with eclampsia, all the subjects had narrowing of the cerebral vessels6. A prospectively designed study of magnetic resonance angiography in nine women with severe pre-eclampsia found vascular narrowing in six5.

In their prospective study that included magnetic resonance angiography in separate groups of women with pre-eclampsia and eclampsia, Morriss et al.7 found that none of their subjects had visible vasospasm of the vessels of the circle of Willis. However, they did not study their women antepartum and, in some women, treatment such as magnesium sulphate and other drug therapies had already been commenced. This may therefore have obscured changes that would otherwise have been present in the more severe cases.

The results of this study show an increase in the NAA/choline ratio in the brain during healthy pregnancy associated with a decrease in choline. In pre-eclampsia, there was a lower NAA/choline ratio compared with healthy pregnancy together with increased choline. By three months postpartum, the results for both healthy pregnancy and pre-eclampsia were similar to those in the nulliparous non-pregnant group. The findings in women with pre-eclampsia are similar to those outside pregnancy. This could suggest that a normal process of adaptation has not occurred in women who develop pre-eclampsia. However, a review of the literature allows an alternative argument to be constructed, namely that these spectroscopic findings reflect cerebral ischaemia.

Studies of animal neuronal cells in culture suggest that NAA is present only in neurones and not in astrocytes or oligodendrocytes and is absent in glial tumours in human brain16,17. In 1H magnetic resonance spectroscopy, it is therefore considered to be a marker of neuronal integrity. However, there is uncertainty regarding which molecules are measured by the choline peak. Studies correlating in vivo proton magnetic resonance spectroscopy in brain tumours with in vitro studies after removal of the lesions have concluded that the choline peak reflects the concentration of the water-soluble choline-containing compounds phosphocholine, glycerophosphocholine and free choline and is a marker of tissue cellularity18,19. However, there are unpredictable changes in the concentrations of choline and acetylcholine after death20 and it is unclear whether similar changes occur after surgical removal of tissue. It is also known that the total concentrations of acetylcholine, phosphocholine, glycerophosphocholine and free choline in the brain are less than 1.6 mmol and this is unlikely to account for the large peak seen on 1H magnetic resonance spectroscopy. It is, therefore, possible that the choline peak represents total brain choline stores20. Why, then, are the choline stores reduced in healthy pregnancy? It may be that this is a reflection of the increased requirement for choline by the fetus.

Choline is obtained through the diet and most foods contain significant quantities of choline21. It is transported into all cells via a low-affinity transport system. Choline entering the cell through this mechanism is coupled to the synthesis of phosphotidyl choline. The fetus derives choline from maternal blood by transfer across the placenta. The placenta is able to store large amounts of choline in the form of acetyl choline, which may act as a reservoir to ensure delivery of choline to the fetus21. There is evidence that these placental stores result in depletion of maternal choline stores. Studies have shown that there are decreased concentrations of choline in the blood of pregnant women22 and that the concentrations of phosphocholine and glycerophosphocholine in the liver of pregnant rats are reduced compared with non-pregnant controls23. Choline is required by the fetus for the synthesis of membrane lipids throughout the body. Large amounts are required by the developing brain, both for membrane synthesis and for the synthesis of acetylcholine. Phosphorus and 1H magnetic resonance spectroscopy studies of the human brain during early postnatal life show very large peaks corresponding to phospholipid precursors and choline containing compounds that decrease progressively during childhood24,25. It is therefore likely that the reduced level of choline in the maternal brain in pregnancy, which we have measured using 1H magnetic resonance spectroscopy, is a reflection of decreased choline concentrations in maternal organs as a result of high fetal demands.

Previous studies using proton magnetic resonance spectroscopy have shown that, in areas of cerebral infarction, there are increased concentrations of lactate, decreased concentrations of NAA and decreased or unchanged concentrations of choline8–11.

Sengar et al.6 studied 10 patients with eclampsia and showed a decrease in the NAA/choline and NAA/creatine ratios compared with healthy age-matched non-pregnant women in areas of the brain where there was hyperintensity on T2 weighted magnetic resonance imaging. This was mainly due to a decrease in the NAA peak. There was also a significant lactate peak. The results of this study suggest that there is neuronal loss in the areas of T2 hyperintensity in eclampsia and that this may be due to ischaemic injury.

Van der Grond et al.12 used proton magnetic resonance spectroscopy to investigate the differences in patients with unilateral stenosis of the internal carotid artery, bilateral stenosis of the carotid artery or unilateral occlusion of the carotid artery who had not suffered a significant cerebral infarct. They compared these patients to a control group of subjects without cerebrovascular disease. In this study, the results showed a decrease in the ratio of NAA/choline in the symptomatic hemisphere of all three groups of patients but with the decrease more marked in those with occlusion of the artery. They also found a high incidence of cerebral lactate in their experimental groups. The decrease in the NAA/choline ratio was caused mainly by an increase in the choline concentration with the NAA remaining largely unchanged.

In our study, we have shown no difference in the NAA peak in healthy pregnant women or in women with pre-eclampsia when compared with non-pregnant controls. This suggests that there is no loss of neuronal integrity in pre-eclampsia. There were also no significant lactate peaks, implying that there is sufficient circulation to the cells within the brain to provide adequate oxygen and remove the products of anaerobic metabolism. However, we found a significantly higher choline in the group with pre-eclampsia compared with the healthy pregnant women. The finding of increased choline is similar to that of van der Grond et al.12 in patients with cerebral ischaemia caused by carotid artery stenosis. As discussed above, there is uncertainty as to which compounds are represented by the choline peak but it is likely that it reflects water-soluble choline-containing compounds to a greater extent than lipid-soluble molecules. Animal studies have shown that, in areas of reduced cerebral blood flow causing impaired function without permanent damage, large amounts of the water soluble compounds glycerophosphocholine, phosphocholine and free choline, are the result of degradation of membrane phosphotidylcholine and sphingomyelin26. We suggest, therefore, that the increase in the choline peak in 1H magnetic resonance spectroscopy in areas of relative cerebral ischaemia is a reflection of this process.

We therefore conclude that there is a lower level of choline in the brains of pregnant women compared with their non-pregnant counterparts and that this may be a reflection of reduced choline stores throughout the body because of demands made by the fetus. Our findings of higher choline with an equivalent level of NAA in pre-eclampsia compared with healthy pregnancy are similar to those in areas of reduced cerebral perfusion and relative cerebral ischaemia in patients with carotid stenosis. The finding of Sengar et al.6 of lower NAA together with high lactate in eclampsia is similar to the findings in areas of cerebral infarction and is suggestive of neuronal loss, probably as a result of cerebral ischaemia. Since the findings in eclampsia are similar to those in cerebral infarction and those in pre-eclampsia are like those of cerebral ischaemia without infarction, we suggest that this is indicative of a similar relative reduction in cerebral perfusion in pre-eclampsia, probably due to cerebral vasospasm, and that this becomes more severe in eclampsia. The relative ischaemia in pre-eclampsia is not severe enough to cause loss of neurones or build up of lactate, but is enough to cause membrane degradation and release of choline containing compounds.

Differences were also detected in the NAA/creatine ratio, with an increase towards the end of pregnancy and a decrease postpartum. There was also a significantly lower NAA/creatine ratio in pre-eclampsia compared with normal pregnancy associated with an increased creatine peak. There is little evidence regarding the significance of changes in creatine. The ‘creatine’ peak represents the concentration of both creatine and phosphocreatine. Phosphocreatine is a reservoir for high energy phosphates in neurones and creatine is produced when this energy is utilised. As the level of phosphocreatine decreases, the level of creatine rises. The ‘creatine’ peak is therefore an indicator of the total energy potential of the tissue but does not represent whether or not energy has been utilised or is available for use. The concentrations of creatine have been found to be higher in glial cells than in neurones27. Studies in patients with cerebral infarction have found that the creatine is decreased8,10,28. In their study of patients with carotid stenosis, van der Grond et al.12 found no significant difference in the NAA/creatine ratio. It is therefore difficult to explain the findings of the present study based on the available evidence.

In conclusion, visible vascular narrowing on magnetic resonance angiography is not a feature of mild to moderate pre-eclampsia may represent severe disease where it is present. The findings on magnetic resonance spectroscopy in pre-eclampsia are similar to those in women who are not pregnant, but are also found in patients with cerebral ischaemia associated with carotid stenosis without cerebral infarction. This suggests that there may be cerebral ischaemia in pre-eclampsia.

Ancillary