If we know so much about preeclampsia, why haven't we cured the disease?

Abstract

Preeclampsia has been recognized for at least 100 years. In the last 20 years, the consideration of the disorder as more than simply hypertension in pregnancy has led to an explosion in knowledge about preeclampsia pathophysiology. It is now evident that for most cases of preeclampsia, the root cause is the placenta. Relatively reduced placental perfusion leads to inflammation, oxidative stress, and endoplasmic reticulum stress, which converge to modify maternal physiology, with endothelium an important target. Although preeclampsia is characteristically diagnosed in the last third of pregnancy, it is evident that many of these pathophysiological changes can be detected long before clinically evident disease. Furthermore, it is evident that the “maternal constitution,” including genetic, behavioral, and metabolic factors, influences the maternal response to the abnormal placentation of preeclampsia. These insights would seem to provide a guide for the prediction of the disorder in early pregnancy, along with targets for intervention. However, this has not been the case. Predictive tests guided by this knowledge do not predict well and several interventions guided by the expanded understanding of pathophysiology do not prevent the disease. We propose that these failures are secondary to the fact that preeclampsia is more than one disorder. Further, we suggest that future progress toward prediction and prevention will require research guided by this concept.

1. Introduction

The adverse pregnancy condition, preeclampsia, has been recognized as a significant contributor to maternal and fetal mortality and morbidity for over 100 years. In the late nineteenth century it was recognized that eclampsia, which had been considered a convulsive disorder of pregnancy for almost a thousand years, was preceded by increased maternal blood pressure and proteinuria, hence preeclampsia. Not long after these observations, it was evident that even without seizures, preeclampsia indicated a potentially fatal disease for mothers and babies (Chesley, 1978).

Early research in preeclampsia was heavily influenced by the original, serendipitous observations of increased blood pressure and proteinuria. Most research was directed at attempts to understand solely the high blood pressure and renal dysfunction. Not surprisingly, little progress was made in understanding the syndrome. In the last 20 years, the recognition that preeclampsia is a multisystemic syndrome that is associated with increased inflammatory activation (Redman et al., 1999), endothelial dysfunction (Roberts et al., 1989), an abnormal balance of angiogenic and antiangiogenic factors (Maynard and Karumanchi, 2011), and profound metabolic changes (Hubel, 2006), has redirected preeclampsia research. As a result, our understanding of the pathophysiology of this disorder has increased strikingly. Furthermore, there has been a substantial increase in the understanding of the process of trophoblast invasion, long recognized as a key feature of preeclampsia (Kaufmann et al., 2003).

Ordinarily, as understanding of the pathophysiology of a disease increases, the ability to predict, prevent, and treat the disease will increase accordingly. This has not been the case with preeclampsia. The inability to treat preeclampsia is not too puzzling since the pathological changes present in women with overt preeclampsia strongly indicate that once clinically evident, the process is not reversible. However, despite an apparently in-depth understanding of the pathophysiology of preeclampsia, we have not been able to reliably predict or prevent the disease. This stimulates the title of this manuscript, a statement made by noted preeclampsia investigator, Joey Granger. “If we know so much about preeclampsia, why haven't we cured the disease?”

In this manuscript we will review the current concepts of the pathophysiology of preeclampsia, explore what is “wrong” with these concepts, and attempt to assess how we should move forward in our study of this enigmatic condition.

2. Current concepts of the pathophysiology of preeclampsia

A useful model for understanding the pathophysiology of preeclampsia has been to consider the disorder with a two-stage model (Gammill and Roberts, 2007, Sargent et al., 2006). In the two-stage model (Figure 1) Stage 1 is inadequate placental perfusion and Stage 2 is the maternal syndrome resulting from inadequate placental perfusion. A crucial question in preeclampsia research is, what links the two stages?

Figure 1. The two-stage model of preeclampsia.

It is proposed that the initiating insult in preeclampsia is a relative reduction of the perfusion of the placenta, Stage 1. This insult leads to the generation of materials (arrow) that act on the maternal constitution to result in the maternal pathophysiology of preeclampsia, Stage 2. Reprinted from Roberts & Hubel (2009), with permission from Elsevier.

2.1 Stage 1: Inadequate placental perfusion

There is abundant evidence that inadequate maternal blood flow to the placenta is present in preeclampsia. This concept is supported by indirect evidence of increased resistance in the distal branches of the uterine artery by Doppler velocimetry (Aardema et al., 2001). Medical conditions associated with microvascular disease (e.g., hypertension and diabetes) also increase the risk of preeclampsia. The concept of inadequate rather than reduced placental perfusion is supported by the association of preeclampsia with obstetrical conditions characterized by increased trophoblastic mass, such as multiple gestations and hydropic placentas. In these settings, it is posited that even normal placental perfusion will not be adequate to supply the increased trophoblast mass. Also, although animal models of preeclampsia are not perfect, there are several experimental manipulations to reduce placental blood flow in several species that will produce a preeclampsia-like syndrome (McCarthy et al., 2011).

The most common cause of inadequate perfusion is failed remodeling of the maternal spiral arteries perfusing the intervillus space. These findings are illustrated in Figure 2 (Parham, 2004). In non-pregnant women, the terminal vessels perfusing the endometrium, the spiral arteries, are similar to small muscular arteries in other parts of the body, with endothelium, smooth muscle, and an elastic lamina. These vessels, if they underlie the placenta, undergo striking remodeling during pregnancy. The diameter of the terminal portion of the artery increases dramatically. The elastic lamina and smooth muscle are no longer present (Brosens et al., 2011). The cells that line the vessel do not appear to be endothelium, but rather fetal trophoblasts that have modified to an endothelial phenotype (Zhou et al., 1997). Importantly, this process extends to the inner third of the myometrium. There are obvious physiologically relevant consequences of this remodeling. The increased terminal diameter has led to the conclusion that there must be a large increase in the volume of flow. However, since dilatation affects only the terminal portion of the vessels, Graham Burton points out, on the basis of modeling studies, that volume flow is increased only about two-fold (Burton et al., 2009). This is very much less than would be predicted by Poiseuille's law, which states that flow increases with the fourth power of the radius if the entire vessel was remodeled.

Figure 2. Failed spiral artery remodeling in preeclampsia.

During pregnancy, the spiral arteries that underlie the placenta, which in nonpregnant women are typical, small muscular arteries containing smooth muscle and an inner elastic lamina, undergo physiological remodeling. This includes terminal dilatation, the loss of the internal elastic lamina, and the loss of smooth muscle. This change extends into the inner third of the myometrium, resulting in the loss of a condensation of vascular smooth muscle near the myometrial decidual junction. This “sphincter” is proposed to be responsible for terminating blood flow at the time of menses. In preeclampsia, this process is not complete. The terminal dilatation is not as extensive and the removal of smooth muscle is not complete and does not extend beyond the decidua, leaving the functional vascular sphincter intact. Reprinted with permission. Parham 2004).

The major impact of terminal dilatation is a reduction in the velocity of blood perfusing the intervillus space. This is an important modification given that blood flow to the uterus increases dramatically during pregnancy (from 0.1 to 10% of cardiac output) (Thaler et al., 1990). If terminal dilatation did not take place, the velocity of the blood perfusing the placenta would be 1–2 m/s, but with remodeling, the flow is reduced to 10 cm/s (Burton et al., 2009).

The loss of smooth muscle is also proposed to be physiologically relevant, preventing humoral or neural regulation of vascular tone in these vessels. In preeclampsia, this remodeling is not complete. Vessels are not as dilated terminally and smooth muscle remains in some vessels, allowing, at least theoretically, humoral or neural modification of vascular tone (Brosens et al., 2011). Another important difference is that the modification that does occur in preeclampsia does not extend into the myometrium. Burton posits that this is quite important, citing the presence of a condensation of smooth muscle at the endometrial and myometrial junction in non-pregnant women, which is eliminated in normal, but not preeclamptic pregnancies. The persistence of this structure, which is proposed to serve as a sphincter that terminates bleeding at the time of endometrial shedding in non-pregnant women, allows further modification of blood flow to the placenta in preeclampsia (Burton et al., 2009).

It is clear that this physiological modification is determined by trophoblast invasion into and around the spiral arteries. This process is compromised in preeclampsia. Regulation of the interaction between fetal and maternal tissues demands unique immunological mechanisms. Thus, syncytiotrophoblast does not contain HLA antigens (Blaschitz et al., 2001). HLA is present on invasive trophoblasts, but is minimally heterogeneous with HLA-G, the predominant antigen (Blaschitz et al., 2001). The detail of these interactions is a subject of intense study, with evidence emerging that combinations of decidual and trophoblast antigens dictate the interaction (Hiby et al., 2004).

Another long-standing observation is that the likelihood of successful implantation is determined by the degree of exposure to paternal antigen. The best example of this is the increased likelihood of preeclampsia in primiparous pregnancies that have not been exposed to fetal antigen by the fetal maternal transfusion that accompanies pregnancy and especially delivery. In addition, the use of barrier contraception (reducing paternal antigen exposure) increases the risk of preeclampsia. Similarly, reduced time between first coital exposure with the father of the child and conception increases the risk of preeclampsia (Robillard et al., 1999).

In preeclampsia, the consequences of inadequate perfusion and the potential for blood flow to the intervillus space to be modified are intermittent hypoxia and the generation of oxidative stress. This has a myriad of down-stream effects, including the release of antiangiogenic proteins, as well as activation of inflammation (Redman et al., 1999). It is likely that changes in the physical characteristics of the syncytiotrophoblast due to these processes, as well as mechanical injury from the greater velocity of blood flowing into the intervillus space, lead to fragmentation of the syncytiotrophoblast surface. Messages carried by humoral materials from the placenta and onward, and in the syncytiotrophoblast fragments, are the likely generators of Stage 2, the maternal syndrome (Gammill and Roberts, 2007).

2.2 Stage 2: The maternal syndrome

It is evident that hypertension and proteinuria, the two diagnostic criteria for preeclampsia, constitute a very small part of the pathophysiology of the syndrome (Roberts and Redman, 1993). Pathological examination of the organs of women with preeclampsia presents a recurring theme of hemorrhage and necrosis, which is consistent with reduced organ perfusion (Roberts, 2009). Pathological examination also reveals a finding present in no other form of hypertension, glomerular endotheliosis (Spargo et al., 1959). This change largely involves the endothelium of the glomerular capillary, suggesting that endothelium may be an important target in preeclampsia. Pathophysiological findings with overt preeclampsia support the reduced organ perfusion indicated by pathological findings (Roberts, 2009). There is reduced perfusion to virtually any organ in women with preeclampsia.

Studies of pathological or pathophysiological changes in women with manifest preeclampsia have minimally informed our understanding of preeclampsia. The profound changes in these sick women preclude discriminating cause and effect. Our understanding of the relevant pathophysiological changes in preeclampsia has been advanced greatly by studies of cohorts of pregnant women who later developed or did not develop the syndrome. In these women, it is evident that there are characteristic pathological and pathophysiological changes present long before clinically evident preeclampsia. Thus, groups of women who will later develop preeclampsia demonstrate several phenomena that would reduce organ perfusion months before clinically evident preeclampsia. These include vasoconstriction secondary to increased pressor sensitivity (Gant et al., 1973), activation of the coagulation cascade with microthrombi (Kobayashi et al., 1999), and loss of fluid from the intravascular space (Campbell and Campbell, 1983) that leads to a reduction in the circulating blood volume. Interestingly, all of these are consistent with endothelial dysfunction (Roberts et al., 1989). The concept that endothelial dysfunction may be a central component of the pathophysiology of preeclampsia has thus received extensive support, with evidence of this process antedating evident preeclampsia (Roberts and Lain, 2002).

2.3 Maternal fetal interactions in preeclampsia

An important question is whether inadequate placental perfusion is sufficient to cause preeclampsia? It is not. Pregnant women with growth restricted infants and one third of women with preterm birth manifest the same abnormal implantation and failed vascular remodeling present in preeclampsia (Brosens et al., 2011). This has led to the conclusion that the result of the inadequate perfusion is influenced by the maternal “constitution.” Thus, underlying maternal disease (e.g., diabetes or hypertension), maternal lifestyle (e.g., obesity, activity, or sleep), genetics (e.g., family history), environment (e.g., air pollution), and other maternal characteristics influence a woman's response to inadequate perfusion and determine whether or not she will develop preeclampsia (Roberts and Von Versen-Hoeynck, 2007). An additional important consideration is that these constitutional factors are influenced by the profound physiological changes accompanying normal pregnancy. One physiological modification that has been recognized in the last 10 years is that pregnancy is accompanied by increased activation of the inflammatory response (Borzychowski et al., 2006). This is further accentuated in preeclampsia and could influence the endothelial and metabolic dysfunction involved in preeclampsia.

There are important implications for the hypothesis that focuses on the importance of maternal–fetal interactions in preeclampsia. First, it provides a target for preeclampsia prevention, both before and during pregnancy. Second, this would explain the increased risk that women with a history of preeclampsia have of cardiovascular disease later in life (Bellamy et al., 2007). Many of the same factors that increase the risk of preeclampsia also increase the risk of cardiovascular disease later in life. Also, it implies that inadequate uterine perfusion may not only be insufficient to cause preeclampsia, but may also not be necessary for preeclampsia. With a combination of maternal factors and inadequate placental perfusion determining outcome, preeclampsia could occur with very reduced placental perfusion and almost no maternal contribution or conversely with minimally reduced perfusion and a profoundly abnormal maternal “constitution”. One could envision that in some extremely sensitive women the stresses of even adequate placental perfusion could result in the maternal syndrome.

2.4 Linkage of the two stages

The “Holy Grail” of preeclampsia research is to identify the linkage of inadequate placental perfusion to the maternal syndrome. For years preeclampsia investigators sought the existence of unique agents in the blood or placentas of women with preeclampsia that might provide this link. Blood and placentas from preeclamptic women were extracted and injected into several different experimental animals, with attempts largely directed at increasing blood pressure and urinary protein excretion. The results, sometimes positive, were universally explained by reactions to foreign proteins or contamination (Chesley, 1978). It is now conclusively established that there are no mysterious toxins in the blood of women with preeclampsia. Interestingly, the endpoint for the earlier studies of preeclampsia, vasoconstriction and hypertension, also appear to be unrelated to increases in any circulating pressor agent, unique or endogenous. Women with preeclampsia, and prior to clinically evident disease (Gant et al., 1973), manifest increased sensitivity to any pressor agent (Talledo, 1966; Talledo et al., 1968). This is consistent with endothelial injury/activation. However, the increased inflammation, endothelial dysfunction, and other aspects of maternal pathophysiology resolve toward pre-pregnant levels after pregnancy and delivery of the placenta. Thus, it appears that placental products, likely increased because of hypoxia and/or oxidative stress, act on a maternal constitution of varying sensitivity to cause the maternal syndrome of preeclampsia. Candidate molecules include antiangiogenic factors (Maynard and Karumanchi, 2011), cytokines (Kalinderis et al., 2011), and syncytiotrophoblast particles or the contents of these structures (Goswami et al., 2006). An extremely attractive “common denominator” for many of these changes is oxidative stress (Raijmakers et al., 2004). Oxidative stress occurs when reactive oxygen species (ROS) are increased beyond the buffering capacity of the organism. There is abundant evidence of oxidative stress and it could both contribute to the release of the several linkers cited and be accentuated by these linkers (Redman and Sargent, 2005).

3. The translation of these current concepts to clinical care

With the wealth of information that has been accumulated in women prior to clinically evident preeclampsia, and the use of these data to support hypotheses relevant to the pathophysiology, it would seem that we are poised for great advances in the management of the disorder. As stated previously, it does not seem likely that treatment at the time of the disease would be more than palliative because of what appear to be irreversible pathological changes. Thus, most efforts have been directed at prevention (which may in fact be early treatment since many of the pathophysiological changes of preeclampsia antedate clinical disease by months). With the low incidence of preeclampsia, it is also evident that enriching any population for whom we plan preventive therapy with predictive testing would be most useful. Therefore, the accumulating data demonstrating pathophysiological changes prior to preeclampsia have guided the search for clinically useful predictors. The translational targets of the pathophysiological targets are prediction and prevention of preeclampsia.

3.1 Predictors do not predict (adequately)

The relationship between preeclampsia and abnormal trophoblast invasion and subsequent deficient spiral artery remodeling has prompted tests that assess the usefulness of trophoblast markers to predict preeclampsia. There are a myriad of these markers that have been tested and many of these demonstrate statistically significant differences between women who subsequently do or do not develop preeclampsia (Grill et al., 2009). However, when predictive values are tested, they are universally too low to be clinically useful. A study performed by the NICHD Maternal Fetal Medicine Network examined trophoblast and angiogenic markers at approximately 12 weeks’ gestation in 174 low-risk nulliparous women who later developed preeclampsia and 509 who did not. There were significant differences between the two groups in the concentration of the trophoblast markers ADAM-12, PAPPA, and the angiogenic factor Phosphatidylinositol-glycan biosynthesis class F (PlGF) in early gestation (Myatt et al., 2012). However, the area under the receiver operator curve (ROC) for these analytes individually was less than 0.62, and when all of the analytes were combined and examined with clinical findings found to be associated with later preeclampsia, the area under the curves was 0.73. This resulted in a sensitivity of 55% at a fixed specificity of 80%. In a population with a prevalence of preeclampsia of 5%, this yields a positive predictive value of 13% and a negative predictive value of 97%. Syncytiotrophoblast particles were not assessed for predictive value in the NICHD study. The few studies on these particles prior to clinical disease are not adequate to judge predictive usefulness, although the studies are encouraging. However, even in these studies these microparticles were only greater than controls in early pregnancy, and not at term (Goswami et al., 2006). Although there are a few exceptions (Poon et al., 2009), this has been the usual outcome of first-trimester predictor studies. Thus far, there is no compelling evidence for the usefulness of these predictors. Similarly, there has been much excitement about angiogenic and antiangiogenic factors measured in the second trimester, but again the predictive value is low (Lapaire et al., 2010).

3.2 Preventive agents do not prevent preeclampsia (well enough to be useful)

The two-stage model of preeclampsia suggests potential targets for prevention. The maternal factors predisposing to preeclampsia or the linkage of the two stages could be targeted. Most efforts have been directed at the linkage. The most extensively studied of the linkages is inflammation, with the use of low-dose aspirin from early pregnancy onward. Although the original rationale for aspirin therapy was to modify prostaglandin production toward a more favorable vasodilator/vasoconstrictor, anticoagulant/procoagulant ratio, by current concepts, it is likely that the anti-inflammatory effect of aspirin is most relevant. The history of studies to reduce the occurrence of preeclampsia with low-dose aspirin provides a prototype for the results of later intervention for preeclampsia with other agents. In the early 1990s, several small studies had documented the successful prevention of preeclampsia by aspirin therapy. A metanalysis in 1991 (Imperiale and Petrulis, 1991) concluded that preventive aspirin treatment reduced the risk of both preeclampsia and cesarean section by two thirds, and reduced the risk of low-birth-weight infants by almost half. These initially encouraging results were evaluated in larger multicenter studies with the finding that there was no evidence of the efficacy of aspirin in reducing preeclampsia or its complications (Sibai et al., 1993; CLASP, 1994) The studies continued with largely negative results until eventually more than 35,000 women had taken aspirin in randomized controlled trials. With this large sample, a statistically significant beneficial effect can be demonstrated by meta-analysis. However, the number to treat to prevent preeclampsia or adverse outcomes is too low to justify therapy except in very high-risk subjects (Askie et al., 2007).

More recently, oxidative stress was explored as a target linkage. The story is quite similar to that of aspirin. In 1999, a small study on antioxidant therapy for women at high risk of preeclampsia (largely abnormal uterine artery velocimetry) resulted in very encouraging findings (Chappell et al., 1999). The study adopted a treatment strategy used previously in studies attempting to prevent cardiovascular disease. This treatment strategy included the administration of 400 IU of Vitamin E and 1000 mg of Vitamin C. The investigators demonstrated that there was a significant reduction in the incidence of preeclampsia and that treatment was accompanied by evidence of reduced oxidative stress and endothelial injury. Once again, this was followed by large multicenter studies of high- and low-risk women (Roberts et al., 2010; Rumbold et al., 2006; Poston et al., 2006). None of the follow-up studies, which used therapies identical to that used in the smaller study, were successful.

Another approach to prevention, the use of calcium supplements, resulted in similar small and larger study discrepancies, but also provided important insights. It was recognized from epidemiological studies that in populations with similar nutritional composition, except for the intake of calcium, populations with a low calcium intake had a high prevalence of preeclampsia and populations with a high calcium intake had a low prevalence of preeclampsia (Villar et al., 1983). This led to several small studies, largely from low and middle income countries, showing the success of supplementation with 1.5 g of calcium daily. In a 1996 meta-analysis, it was concluded that calcium was an inexpensive approach to reducing the rate of preeclampsia by 60% (Bucher et al., 1996). Once again, a larger multicenter study sponsored by the NIH in the United States followed, and again there was no evidence of the efficacy of calcium supplementation in reducing the rate of preeclampsia (Levine et al., 1997). An attempt to determine an explanation for the discrepancy reached an important conclusion. Although the drug doses, compliance with usage, and to a large extent, diagnostic endpoints, were similar in the large NIH study and the previously reported smaller studies, the populations treated were dramatically different. All of the small studies came from low- and middle-income countries, while the NIH study was carried out in the USA. It seemed very likely that the baseline calcium intake was different in the two populations, with intake being much lower in the low- and middle-income countries. The WHO, stimulated by this concept, assessed calcium intake in several developing countries and carried out a study supplementing calcium in populations with a low calcium intake (Villar et al., 2006). Although the prevalence of preeclampsia was not reduced, severe outcomes were reduced. The rate of eclampsia and severe hypertension was reduced by 30% and a composite outcome of severe maternal and neonatal morbidities was reduced by 25%. This observation that a subset of women with preeclampsia might benefit from preventive strategies has important implications.

4. Why does prevention not prevent and prediction not predict?

When large studies do not demonstrate efficacy for a therapy, while small studies do, the most obvious explanation is publication bias. That is, small studies that are effective are published, while small studies in which intervention was not successful are not, and may not even be submitted for publication. It is also possible that the failure to identify predictors or preventive agents is because we have not chosen the right link between Stage 1 and Stage 2 or that treatment was too early, or too late, or too little, or too much.

There is, however, another important possibility. It is well recognized that the diagnostic criteria for preeclampsia, hypertension, and proteinuria are nonspecific. The clinical presentation of preeclampsia is also heterogeneous, with some women going from initial diagnosis to a life-threatening disease in hours to days, while in others the clinical condition is quite stable with very slow progression. Despite the concept that Stage 1 of preeclampsia is secondary to reduced placental perfusion, only one third of pregnancies affected by preeclampsia are complicated by fetal growth restriction (Eskenazi et al., 1993). Further, an excess of growth-restricted fetuses is only present in a subset of pregnancies affected by preeclampsia that deliver before 37 weeks’ gestation (Xiong and Fraser, 2004). Also, there are other differences in the 10% of pregnancies affected by preeclampsia that deliver early. Although all pregnancies affected by preeclampsia are associated with increased cardiovascular disease in later life, this risk is almost doubled in pregnancies that deliver at term and is increased almost 10 times in women whose pregnancy affected by preeclampsia terminates before 34 weeks’ gestation (Mongraw-Chaffin et al., 2010). A characteristic finding when laboratory findings in women with preeclampsia are examined critically is illustrated in Figure 3 (Hubel et al., 1996, Powers et al., 2005). There is an enormous overlap in values, with some women with preeclampsia having values similar to normal pregnancy. This is true even for the purported linkers of Stage 1 and Stage 2. Not every pregnancy with preeclampsia has an increased concentration of any of the proposed linkers (Figure 3).

Figure 3. The heterogeneity of laboratory findings in preeclampsia.

Scattergrams indicate three well-established pathophysiological findings in preeclampsia. A and B indicate triglycerides and malondialdehyde concentrations before and 24 h after delivery in women with preeclampsia (PE) or without preeclampsia (Norm). C demonstrates the concentration of s-Flt in women with severe or mild preeclampsia and normal pregnancy. The wide scatter of the findings, and the overlapping of the data from women with and without preeclampsia, is typical of findings with many analytes in preeclampsia. A and B reprinted from Hubel et al. (1996), with permission from Elsevier. C reprinted from Powers et al. (2005), with permission from Elsevier.

Thus, another possible explanation for our frustration in preventing or predicting preeclampsia is that we are not dealing with one disease (Roberts and Hubel, 2009). The data strongly suggest that there are several subtypes of preeclampsia. With this explanation, the reason why small studies are effective is because they are from single centers with a homogeneous population, where as large studies are heterogeneous with samples from many centers. The disheartening conclusion from this explanation is that no preventive agent will prevent and no predictor will predict all cases of preeclampsia. The encouraging conclusion is that preventive agents will prevent some cases of preeclampsia and predictors will predict some cases of preeclampsia. However, this latter favorable scenario depends on the recognition of the subtypes of preeclampsia. The importance of appropriately classifying the subtypes of a disease is illustrated by the recognition of Type 1 and Type 2 diabetes. Where would we be in our understanding and treatment of diabetes if we believed that all diabetes was insulinopenic?

5. Subtypes of preeclampsia?

We have alluded to some obvious potential subtypes that warrant studies searching for different pathophysiological bases. Are early and late preeclampsia or preeclampsia with or without fetal growth restriction all one disease? Preeclampsia also seems to primarily affect different target organs in different women. This may be due to varying maternal sensitivity, but it is nonetheless worth exploring similarities and differences in women with preeclampsia primarily affecting the liver, the kidney, the cardiovascular system, or the placenta to search for different pathophysiological subtypes.

It should also be useful to exploit laboratory findings that support different linkages in women with preeclampsia. Is it possible that there are women with antiangiogenic, inflammatory, oxidative stress or endoplasmic reticulum stress as the linkers? We have generated data in our group that support this hypothesis.

We measured the angiogenic factor PlGF across pregnancy in the blood of 50 women who later developed preeclampsia and 250 who did not (Powers et al., 2012). The findings in Figure 4 illustrate the distribution of values. The concentrations of PlGF varied widely, but were significantly lower 7–10 weeks before delivery in women who went on to develop preeclampsia. However, an ROC curve demonstrated an area under the curve of 0.64, and in a population with a prevalence of preeclampsia of 5%, the positive predictive value at a fixed specificity of 80% would be 9.6%.

Figure 4. Phosphatidylinositol-glycan biosynthesis class F (PIGF) concentrations across pregnancy.

PIGF was measured across pregnancy in 50 women who later developed preeclampsia (□) or 250 women who did not develop preeclampsia (◆). Between 15 and 22 weeks of pregnancy, values are significantly lower in women who later developed preeclampsia (p < 0.04). However, the ROC curve for these data had an area under the curve of 0.64, and if a cut off value was selected that resulted in a 20% false=positive rate, the accompanying sensitivity was 40%. In a population in which preeclampsia developed in 5% of pregnancy, the positive predictive value would be 9.5%.

However, if the data were examined longitudinally there were two very clear groups of women who developed preeclampsia (Figure 5). About half of these women had PlGF concentrations less than the 95% confidence interval of subjects with normal pregnancy outcome from 15 weeks’ gestation. This persisted throughout pregnancy, including when they had clinical preeclampsia. The other half of women destined to develop preeclampsia had PlGF concentration similar to women with normal pregnancy outcome, including when they had clinically evident preeclampsia. The low PlGF and high PlGF subjects had very different clinical features. The women with low PlGF had higher blood pressure before 20 weeks’ gestation (prior to clinically evident preeclampsia) and after clinical evidence of preeclampsia, delivered earlier with an excess of preterm birth, and were more likely to have small-forgestational-age infants. We are exploring these exciting observations, which seem to support a low angiogenic and normal angiogenic form of preeclampsia.

Figure 5. Longitudinal pattern of PIGF concentrations.

The data presented in Figure 4 were examined as longitudinal values for individual women. In 27 of the women who went on to develop preeclampsia, (A) the concentration of PIGF was similar to that in women with normal pregnancy outcome, prior to and with clinical preeclampsia. Normal pregnancy findings are illustrated in gray as median (solid line) or 95% confidence intervals (dashed line) in both A and B. In A, the results from women with preeclampsia are in blue (median solid. 95% confidence interval, dashed). The other 23 women who developed preeclampsia manifested a very different pattern (B), with no overlap with normal values, being less than the 95% confidence intervals of normal results from 15 weeks’ gestational age. Results from these women with preeclampsia are in red (Median solid 95% confidence intervals, dashed.) Reprinted from Powers et al. 2012, with permission from Wolters Kluwer Health.

Future research on preeclampsia must be carried out in a manner that allows us to test for (and we hope identify) different preeclampsia subtypes. We have alluded to clinical and epidemiological findings that suggest different subtypes. It is important to look at these groups separately as we probe pathophysiology. However, a problem that must be dealt with is that in many current studies on preeclampsia, obviously heterogeneous groups are combined. Combining all groups of women at increased risk of preeclampsia as a high-risk group is one such error. Should it be surprising that women with an increased trophoblast load with a multiple gestation or diabetes might manifest a different pathophysiological pathway and different response to preventive therapy, and exhibit different predictors from a woman with preexisting hypertension or previous preeclampsia? Similarly, preeclampsia in a first pregnancy, compared with recurrent preeclampsia in subsequent pregnancies, has very different long-range implications. Is preeclampsia in lean and obese women the same disease? Although it is evident that new onset hypertension without proteinuria (gestational hypertension) may sometimes, especially when accompanied by other systemic findings, represent the same syndrome as preeclampsia, this is clearly not always the case.

The approach that is indicated by these questions and facts is one that is nearly opposite to our current approaches. Whereas women with diverse findings are currently combined in studies of preeclampsia, the strategy guided by the concept that preeclampsia is a disorder with several subtypes argues that any obvious or not so obvious differences should be considered a different subtype until proven otherwise. We should also exploit differences in laboratory studies in groups of women with preeclampsia. The tendency to present data as means and medians with variance obscures the fact that laboratory findings in preeclampsia are quite variable. This approach has led to the concept that all women with preeclampsia have a specific pathophysiological finding that only varies by degree of involvement. The actual data show a very different pattern (Figure 3). Somewhere in the evaluation of an analyte of interest (perhaps even with publication), scattergrams should be examined and unique features of outliers explored.

Preeclampsia is a disorder defined by nonspecific signs that presents with extraordinarily diverse clinical and laboratory findings and later life implications. Only when a rigorous research approach is initiated, which acts on the logical assumption that preeclampsia is more than a single disorder, will we successfully translate research findings in order to predict and prevent preeclampsia(s).