Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Harmonization of Data and Biobanks for Preeclampsia Research
Editor's comment : This new chapter is included in recognition of the increasing power of informatics and data analysis that have become evident in the 21 st century. We are no longer constrained to examining tiny “slices of disease” or attempting to satisfy the Ockham's razor approach to simplifying findings to the most parsimonious explanation. Rather, we now have the power to begin to examine interactions of genes, environment, and behavior. This is especially relevant in pregnancy in which we deal with two individuals, mother and fetus, and the even more complex situation when things “go wrong” with one or both of these individuals as occurs with preeclampsia. In addition, many adverse pregnancy outcomes such as preeclampsia are syndromes, defined by findings that were first observed or are most common in the disorder, which have been frequently misinterpreted as the “most important” changes. It is also increasingly evident that preeclampsia is heterogeneous and that deciphering this heterogeneity and determining what are potential therapeutic targets are key to understanding, predicting, and preventing the disorder. However, the power of these analytical strategies mandates large amounts of data to be carefully and consistently collected. This is not possible in a single center and requires collaboration. Furthermore, data must be carefully collected in a manner that allows harmonization. As appealing and useful as these approaches sound, they are associated with substantial challenges. These challenges must be addressed, and this chapter suggests solutions.
Introduction: Why Do We Need to Harmonize?
Limited Progress
Hypertensive disorders of pregnancy, and in particular preeclampsia, remain a major health problem worldwide for mothers and babies, with huge impact on families and societies. The societal and individual costs of these syndromes remain high in high-income countries, but this burden is excessive in low- and middle-income countries. In these settings not only are antenatal and obstetric resources scarce, but also epidemiological studies and time for research are limited.
Intensified antenatal, obstetric, and postpartum care in high resource settings has dramatically improved maternal and neonatal outcomes over the last century. Improvement in outcomes has depended on many advances in maternal and perinatal care in general, but also on systematic and adequate antenatal screening underpin preemptive delivery. In the last 40 years, there have been striking increases in our understanding of the pathophysiology of preeclampsia and the recognition that preeclampsia is more than pregnancy-induced hypertension. Despite this, our management is largely the same as it has been for the last century; observation and timely delivery. Efforts to predict and prevent preeclampsia have had limited practical success despite the advances in understanding. Regardless of improved screening efforts over the last few years, including the use of a Bayes theorem-based model of competing risk in several populations, most preeclampsia cases remain unpredictable, in particular those with late-onset (see Chapter 4 ). Also, prevention by aspirin is currently the prophylactic therapy with best demonstrated effect, but it has not proven useful in women with chronic hypertension deemed at high risk for preeclampsia.
Preeclampsia May Be More Than One Disease
Intensive research efforts over the last 2–3 decades now recognize that hypertensive disorders of pregnancy, and particularly preeclampsia, are multisystemic syndromes, including important aspects of immunology, placental dysfunction, vascular inflammation (including endothelial dysfunction), angiogenic factors, and cardiovascular effects. Not surprisingly this complex array of pathophysiologies is associated with highly variable clinical presentations. Preeclampsia usually presents after 20 weeks' gestation and may present before, during, or after delivery. Eclampsia has been documented as early as 16 weeks and as late as 22 days after delivery. In keeping with this heterogeneity, laboratory findings are remarkably variable. These include changes that have been proposed as relevant to pathophysiology including oxidative stress and dysregulated production of angiogenic factors. Although these variable findings are present in preeclampsia at any stage of gestation, they may be most beneficial in distinguishing early- from late-onset preeclampsia, especially if separated based on delivery prior to or after 34 gestational weeks. Preeclampsia beyond 37 weeks is not usually associated with an increased risk of fetal growth restriction and the severity of maternal disease is usually less. This difference is also characteristic of long-term outcomes with later-life cardiovascular disease risk doubled with late-onset preeclampsia but increased almost eightfold with preeclampsia occurring before 34 weeks' gestation.
Challenges of Dealing With Conditions Defined by Syndromes
These differences in what we have previously considered to be one “disease” arise from the fact that preeclampsia is not a disease but a syndrome. A syndrome is an empirical definition of a clinical presentation by a cluster of features. It reveals nothing about pathogenesis. Typically the cluster can vary but there is overlap between different components of the cluster as demonstrated in Fig. 21.1 .
Syndromes are convenient aids for guiding clinical practice. The presence of a syndromic signs can help identify an underlying disease, although their absence cannot exclude the condition. In other regards, syndromes are limited. First there is no gold standard, the best that is possible is a consensus encompassed in different national or international guidelines. These may impose a degree of consistency but not necessarily agreement and may create important biases. Standard indices of test effectiveness, for example, test sensitivities or specificities are always disappointing when applied to a syndrome. In short, syndromes are readily recognized clinically but rarely does their diagnosis identify the nature or cause of the problem.
Many syndromes, including preeclampsia, are indicators of common complex diseases. These arise when polygenic responses interact with variable environmental factors to promote a common clinical endpoint. Preeclampsia is unusually complex in that it involves two genomes, maternal and fetal (placental) as well as external factors. The final pathogenesis is syncytiotrophoblast stress, which results in a syndrome with the interaction of a range of maternal responses to dysfunctional placental–maternal signaling. Syncytiotrophoblast stress is a common convergence point generating the maternal syndrome, but it can be generated by several pathophysiological pathways including poor placentation (early-onset preeclampsia) or placental aging (late-onset preeclampsia), or rarer causes such as infection with parvovirus (mirror syndrome) and likely other pathways.
To cope with the shortcomings of syndromes in clinical practice, it is necessary to be realistic and flexible. We need definitions to guide day-to-day practice and to serve the needs of clinical audit or epidemiological studies. But their unavoidable limitations need to be acknowledged. They are not absolute rules but at best advisory. The clusters of features that comprise the syndrome, as in a Venn Diagram ( Fig. 21.1 ), have overlapping or distinct parts. This inevitably leads to apparent disease variants: such as nonproteinuric or even normotensive preeclampsia; or preeclampsia with our without fetal growth restriction, with or without eclampsia or the HELLP syndrome or other atypical presentations. In truth, these are not disease variants but syndrome variants. In syndromes, heterogeneity is to be expected, it is part of the concept. There are limits to the progress that can be made by refining the definitions, by making them more rigid with tighter rules of inclusion or exclusion, when the pathogenesis is not well understood. Even after redefining uncertainty, one is still left with uncertainty.
For clinical research new strategies are needed. Research investigations have been classically based on the Cartesian concept of linear causality—that A causes B—a change down a straight path. Complex disorders evolve differently, around and across intricate networks, in which it is possible to reach the same outcome by different routes, which may be circuitous. These are best studied by omics, systems biology, and modern computer-based statistics and modeling, where the interconnections are not assumed and the modeling is not used to test a hypothesis but, without prior assumptions, to demonstrate the functions of the network in disease states.
Comprehensive phenotyping is necessary to maximize the usefulness of these methods. In preeclampsia and other obstetric syndromes, phenotyping is especially demanding because these result from three genomes (maternal, paternal, and fetal) and three interacting phenotypes—the mother, the fetus, and the placenta—the placenta being important because, in the last analysis, preeclampsia is a placental disorder. Moreover, the phenotyping should not prejudge the profiles of the condition that are most important; that is, the task of the modeling and systems analyses to discover. Hence the phenotyping information should not be categorized dichotomously but comprise the actual continuous observations as they are recorded—not “fetal growth restriction” but the observed birth weight centile, for example. Such maternal responses (for example, gestational hypertension or proteinuria) are the detectable consequences of pathological placental signaling. At the bedside we see a reflection (comprising maternal responses) of the problem, not the problem itself. The mother acts as a mirror and her particular physiology introduces refractions and distortions. We should not be surprised that different mothers produce different reflections of similar syncytiotrophoblast problems. Therefore, to classify preeclampsia it would be helpful, indeed essential, to have measures of the underlying syncytiotrophoblast stress. Placenta-derived biomarkers, which are responses to hypoxia or other stressors, such as low maternal circulating PlGF protein concentrations, could help to classify preeclamptic pregnancies that are more similar in pathophysiology. Ideally every case should include availability of placental tissue to determine the health of the chorionic trophoblast, both cytotrophoblast and syncytiotrophoblast. As for placental phenotyping, there is no agreed option except that classical clinical histopathology is not enough, since syncytiotrophoblast stress is not well detected in this way.
In summary, future preeclampsia research should be better organized around informatics and modeling.
What Has Hindered Our Ability to Unravel the Complexity of Preeclampsia?
Limitations in the Analysis of Complex Data
For many years we have been forced to constrain studies of the complexity of diseases because of the limited ability to handle this complexity analytically. We have given lip service to the importance of gene–environment interactions that strikingly minimizes the magnitude of the issue. Gene–environment interactions are not the interaction of two factors but rather of thousands. Genetic effects are genomic and epigenomic, involving repressors, and enhancers to mention a few, while environmental effects on physiology range from behavior, toxins, and pollution as well as environmental effects that are beneficial. This is all magnified in pregnancy in which there are two individual genomes and environments, that of the mother and fetus. We have attempted in the past to avoid this complexity by studying a “narrow slice” of a disease. Hypothesis-driven research was considered the epitome of a successful research strategy, looking at a precise (limited) component of a disorder while attempting to minimize or control for confounders by experimental or analytical design. This, despite the fact that it was often evident that unrecognized interactions (confounders) might provide insight beyond the original hypothesis. Progress in bioinformatics now allows us to begin to deal with these complexities. However, although investigators are beginning to recognize the importance and power of these new techniques, there are obstacles to this approach, well illustrated by the frustrations we experience in studying preeclampsia.
There is a lack of standardization in the approach to studying this complex syndrome, including its long-lasting health impacts. This lack of standardization and phenotyping is secondary to the lack of uniform and clear definitions of phenotype subgroups that may have distinct pathogenetic pathways. A major limitation of individual studies is that they have low subject numbers and often small effect sizes, and because of lack of standardization, the findings are less likely to be reproducible. To date, individual research groups and consortia usually have collected data using their own specific methodology, unique definitions, and data capture tools, making comparisons between studies difficult. ,
Among the suggested actions to make more published research results “true,” Ioannidis highlights the need for large-scale collaborative research, replication culture, registration, sharing, reproducibility practices, standardization of definitions and analyses, and improvement in study design standards, as discussed in this chapter. Even large single-center or collaborative studies may lack sufficient power to study “pure phenotypes” among the obstetric syndromes (including preeclampsia) and rely on arbitrarily defined clinical observations (such as blood pressure, birth weight, and gestational age), which may not accurately reflect the underlying pathophysiology. , Also, the majority of global health research data, including the study of preeclampsia, are highly biased to represent high-income countries, where the most severe pregnancy outcomes are less common and often censored by access to antenatal care and obstetric intervention.
You're Reading a Preview
Become a Clinical Tree membership for Full access and enjoy Unlimited articles
If you are a member. Log in here