Research on recurrent pregnancy complications: a clinician's perspective

Authors

  • John C. Smulian

    1. Division of Maternal-Fetal Medicine, Department of Obstetrics, Gynecology and Reproductive Sciences, UMDNJ-Robert Wood Johnson Medical School, New Brunswick, NJ, USA
    Search for more papers by this author

Dr John C. Smulian, Division of Maternal-Fetal Medicine, Department of Obstetrics, Gynecology and Reproductive Sciences, UMDNJ-Robert Wood Johnson Medical School, 125 Paterson Street, New Brunswick, NJ 08901-1977, USA. E-mail: smuliajc@umdnj.edu

Summary

Research on recurrent pregnancy complications is essential to help clinicians provide appropriate counselling and guide the management of patients with a history of adverse pregnancy outcomes. However, recurrence research is complex, in both its execution and interpretation. The paucity of appropriate data sets accessible for study of recurrent pregnancy outcomes presents significant challenges in performing recurrence research. This is further compounded by the different perspectives on recurrence between epidemiologists and clinicians. The interpretation of risk, whether it is absolute risk, relative risk or population-attributable fraction, underlies the often opposing perspectives of researchers and clinicians. Because clinicians acutely feel the need to provide appropriate counselling and management strategies when there has been a previous pregnancy complication, it is necessary that all those involved in research and care for these women work together to address gaps in our knowledge for recurrent pregnancy outcomes. In this way, we can develop a better understanding of disease processes, counsel patients better, design management plans and, ultimately, achieve better outcomes for our patients.

Introduction

History . . . is an analysis of the past in order that we may understand the present and guide our conduct into the future. –Sidney E. Mead

Any complication of a pregnancy has the potential to have significant impact on the health and well-being of both mother and child. Even if health is preserved, the psychological impact can be potentially devastating on the mother and family. The effects of having had an important complication often carry over into a subsequent pregnancy, whether through changes in health, changes in clinical management, or through stress and anxiety. This lingering effect is influenced by the complex interactions of environment and genetics. Because such complications often recur, it is important to examine and focus research on this ‘enriched-risk population’. Research on recurrent pregnancy complications is essential to help clinicians provide appropriate counselling and targeted management plans to patients with a history of adverse pregnancy outcome(s). For those women who have already experienced what may have been an unusual or rare pregnancy complication, it is insufficient to counsel about overall population risks for a next pregnancy.

First, general population risks may or may not apply to recurrence of most complications. Unfortunately, there are limited data to provide refined recurrence risks for most pregnancy complications. Second, patients tend to be sceptical about ‘general’ risk numbers. Counselling someone that the chance is very small of ‘being hit by lightning’ means little if lightning has already struck that individual. Focusing research on recurrent pregnancy complications will help improve a clinician's ability to individualise risk and provide patients with a more honest and accurate assessment of what to expect in the next pregnancy. Importantly, obstetricians want to be able to use this better information to design targeted management strategies for subsequent pregnancies. Nevertheless, the clinical subtleties of obstetric diseases that are prone to recur, highlight the challenges of understanding and individualising risk.

There are a variety of complications associated with pregnancy that have been identified as having a high recurrence risk. Selected examples of these are shown in Table 1. In recent years, it has become clear that all of these conditions have a complex mix of risk factors and aetiologies. The recognition of such heterogeneity within each complication category has made it very challenging to accurately study and quantify risk that relates to recurrence of pregnancy complications. As in most areas of medicine, obstetricians rely on the results of epidemiological research to guide counselling and care when clinical research is not available. This makes it imperative to improve research in the area of recurrent pregnancy outcomes in order to help obstetricians provide accurate information to their patients.

Table 1.  Selected pregnancy complications with increased risk for recurrence
Preterm birthShoulder dystocia
Pre-eclampsiaPostpartum haemorrhage
Fetal growth restrictionPregnancy loss
Placental abruptionGestational diabetes
Placenta praeviaThromboembolism
Fetal malpresentationBirth defects
StillbirthNeonatal group B
streptococcus sepsis

The perils of interpretation of data by clinicians

Much of the quantitative assessment of risk for pregnancy outcomes has been defined by epidemiological research, but clinicians often share a sceptical view of epidemiology. Clinicians generally recognise the need to go beyond the practice of ‘anecdotal medicine’ where personal experience, admittedly biased, is the primary force behind medical decision making. However, population-based research is often viewed as mysterious manipulation of data using sophisticated statistical methodology and terminology that is hard to understand, giving results that are, therefore, suspect. As one of the primary goals of epidemiological research is to understand disease aetiology and only secondarily to address issues of prediction, there is an inherent gulf between epidemiological researchers and clinicians who are looking for ways to improve their clinical practice. For example, a modest relative risk of 2.0 linking a particular risk factor with an outcome may provide important clues to aetiology for a specific disease, whereas it may take a many-fold higher relative risk to make a clinically useful predictor of outcome. However, epidemiologists should be aware that clinicians may place disproportionately more emphasis on weaker degrees of association and change clinical care prematurely due to their biased perspectives when dealing with women who have had a previous pregnancy complication.

The history of a previous complication should focus attention on the possibility of recurrence, but most recurrence risks are poorly quantified, leaving clinicians to their own biases. Some clinicians, in the absence of a clearly quantified risk for recurrence, will minimise risk in counselling their patients so as to not upset them (i.e. ‘Chances are, it won’t happen again'). Others overemphasise the risk to patients in an attempt to reduce potential liability in the face of a future recurrence, such as with counselling a patient with a previous delivery complicated by a shoulder dystocia (i.e. ‘Do you really want to take that chance again?’).

Unfortunately, many statistical concepts common in epidemiology are misinterpreted by clinicians trying to apply research results in their clinical practice. It is not unusual for the average clinician to misinterpret odds ratios and relative risks. A relative risk of 1.5 associated with an exposure represents a 50% increased risk of disease outcome over those not exposed, but this can be mistakenly interpreted as implying that the disease outcome of interest will be reduced by 50% if the risk factor is eliminated. This latter interpretation is one of the population-attributable fractions (aetiological fraction). The pressure to misinterpret epidemiological data has at least some of its origins in the fact that individual clinicians (in this case obstetricians) think in terms of absolute risks. They really want to know what are the chances that a particular patient will develop a particular pregnancy complication. This type of information is important when counselling and managing individual patients and is increasingly needed in this era of liability peril. While our overall ability to individualise risk is relatively poor, for many complications the history of a specific adverse pregnancy outcome is the greatest risk factor for that outcome in a subsequent pregnancy.

Clinicians also have a tendency to ignore that statistical associations do not imply causality and that causality is usually not linear. For example, if 29% of preterm births are associated with bacterial vaginosis (BV),1 some clinicians would have an almost irresistible urge to make a leap to causality and conclude that, ‘If I can treat all my pregnant patients with BV then I can reduce preterm births in my practice by 29%’. Clearly this is an overly simplistic and, more importantly, incorrect extrapolation of data, as this conclusion does not account for a history of infection-related preterm birth in a previous pregnancy, BV colonisation rates in the clinician's practice, patient compliance rates with treatment, organism response to treatment, or proportion of preterm births attributable to infection in that population (and many more). We also know that any association of BV and preterm birth is likely to be mediated by numerous inflammatory pathways, and simply treating BV will have a relatively limited effect on preterm birth.2–4 Nevertheless, many clinicians used early BV data to enthusiastically target counselling and aggressive treatment of BV, before confirmation of efficacy. As time has shown, an aggressive approach to BV is best reserved for women with a previous spontaneous preterm birth in order to prevent recurrence, especially when the previous preterm birth was associated with intra-amniotic infection. Unfortunately, there is still widespread use of antibiotics to treat asymptomatic women with BV who have no history of previous preterm birth.

Misinterpretation of research can lead to both premature changes in health practices as with the BV story, but also lead to highly variable interventions, which introduce uncontrolled heterogeneity into management. Some interventions that were initially greeted with enthusiasm, may actually cause harm, as was the case throughout the 1960s with widespread use of diethyl-stilboestrol therapy to prevent miscarriage.

An example of the complexity of how absolute and relative risks can affect clinical care is demonstrated for women who have had a thromboembolic event (TEE) in a previous pregnancy. The incidence for a primary TEE in pregnancy is estimated to be approximately 8.5 per 10 000 (0.085%).5 For low-risk women with a previous TEE during pregnancy that was associated with a transient risk factor and no detectable thrombophilia, the overall absolute risk for recurrence is relatively small, at approximately 50 per 10 000 births (0.5%). For high-risk women, the recurrence risk is approximately 600 per 10 000 (6%) and anticoagulant therapy is considered standard care.6,7 The overall risk for recurrence is very small in the low-risk group (approximately 99.5% will have no TEE recurrence) and interventions with full anticoagulation have not been proved to have substantially improved outcomes in this group.6,7

Interestingly, some clinicians (and authorities) provide the interpretation that the low overall recurrence rate combined with potential anticoagulant therapy complications does not warrant prophylactic use of anticoagulants until after delivery. Others contend that the nearly sixfold increased risk (8.5–50 per 10 000 births) is the relevant number that is sufficiently elevated to warrant therapy. To put it in appropriate perspective, the concept of number needed to treat so as to prevent a recurrence of TEE can influence the clinical approach to counselling. Keep in mind that the patient may have a different interpretation of the risk numbers based on the specific complications and severity of illness surrounding her previous TEE experience. Her experience with either the TEE or the therapy may influence her perspective on prevention strategies in a subsequent pregnancy.

Clinical considerations in recurrence research

Preterm birth is a useful example to highlight some of the complexity of recurrence research. Recurrence risk for preterm birth has been estimated to be as high as 45%. With the reported increasing rate of preterm birth across the US to more than 12% and the high recurrence risk, there is a focused effort to understand and, ultimately, prevent preterm birth. Very recent data from Ananth et al.8 have suggested that a significant portion of the temporal increase in overall preterm birth in the US was associated with a temporal increase in medically indicated preterm births. If indeed the rise in preterm births is primarily due to medically indicated interventions, then the approaches towards understanding risks and preventions should be targeted towards understanding the aetiologies of all of the pregnancy complications that result in medically indicated preterm birth. For instance, if the preterm birth was related to pre-eclampsia, then interventions proposed to prevent spontaneous preterm birth, such as weekly administration of 17-hydroxy progesterone, are unlikely to be effective, whereas interventions to prevent the recurrence of pre-eclampsia may be a better strategy to prevent the recurrence of preterm birth specifically in these women.

Interestingly, a growing number of recent studies have suggested that the presence of one obstetric complication in an index pregnancy presents a higher risk for both recurrence of that specific problem and occurrences of other complications previously thought to be unrelated. Because of this, recurrence research needs to focus on families of conditions related by potential aetiology (i.e. pathways), as opposed to a specific complication itself.9,10 For example, placental conditions may lead to the seemingly different outcomes of fetal growth restriction, pre-eclampsia or placental abruption. The challenge is to identify a mechanism to study the complex interactions of pathophysiology and outcomes.

Unfortunately, patterns of disease that are observed clinically are often difficult to demonstrate statistically. Administrative data sets with linked data across pregnancies are relatively few and the data for examining specific risk factors and interrelated conditions are often ‘sparse’. Related family histories that would give clues to disease pathways, are usually not included and important details of disease states (such as severity) are often not available. Clinical research on recurrence is very important, but runs the risk of being limited by study design to explore other potentially related outcome conditions outside the primary outcome of interest.

A thorough clinician will take an extensive history from a patient that includes all aspects of obstetric, gynaecological, medical, surgical, social and family history. This will be combined with physical examination data to assess a patient's general risks for various pregnancy outcomes. Obstetricians are generally comfortable with distilling this tremendous amount of data into useful information for what they consider reasonable patient management. Existing recurrence risk studies, using administrative data sets, clinical record reviews or data collection questionnaires that estimate risks for recurrence, are usually based on sources that have a relatively limited amount of detail compared with the detail obtained through the physician–patient interaction. Therefore, epidemiologists should be aware that, when obstetricians are presented with epidemiological studies that incorporate limited variables within complex analyses (especially those that adjust for confounders), it may be difficult for a clinician to figure out how the results are clinically significant.

Finally, in clinical practice, obstetricians deal constantly with the patient's perspective when considering how to interpret risk. Patients want to know for any given complication they have already experienced, ‘Will it happen to me again?’ The response to this question from the clinician is most often to answer using a binomial approach; that is, an estimated risk for a particular outcome across a range of 0% to 100%. In reality, what the patient is really asking is for a Bernoulli interpretation of her risk, i.e. is my risk 0% (no) or 100% (yes) for experiencing a particular complication? This is the question that is the hardest to answer. The way that a clinician presents risk often drives a patient's decision making with respect to consenting to follow a particular management plan. It is the challenge for all those performing research into recurrent pregnancy complications to address this in the most direct way possible through refining our understanding of risk for a particular condition.

Clinical bias

Interestingly, clinicians often perceive risk factors differently than the epidemiologist. Obstetricians are constantly looking for ways to improve pregnancy outcomes, particularly for women who have had a previous adverse event. Therefore, the most relevant risk factors for clinicians are those that are potentially modifiable. It is very unappealing and not very productive for the physician–patient relationship to counsel a pregnant woman that her parity or race predisposes her to a pregnancy complication. Therefore, obstetricians are often biased to assign relatively more importance to risk factors that are believed to have interventions, however weak the association may be. This can inadvertently lead to management decisions that are not based on good evidence and will contribute to wider variations in clinical care.

It must also be remembered that the ability to examine recurrent pregnancy complications is often compromised by the clinical management of a second pregnancy after a pregnancy complication. Pregnancies at risk for recurrence of pre-eclampsia may be delivered at the first sign of elevated blood pressure or abnormal fetal growth prior to the development of classical pre-eclampsia. This may result in a preterm birth from the medical intervention with no actual diagnosis of pre-eclampsia. There are no data sets that can reflect this type of clinical management bias. Because of this ‘confounding by indication’, it is often very difficult to derive more precise estimates of risk. Unfortunately, this problem is inherent in studying an area that requires active clinical management.

Conclusion

Ultimately, it is the need to provide precise risks and to design intervention strategies targeted to specific pathophysiologies that should drive an effort to understand recurrence of obstetric complications. Understanding the complex interactions among the various pregnancy complications and their influence on recurrence risk is essential for providing clinicians with reasonable data with which to counsel patients and to manage pregnancies. Nevertheless, researchers should be aware of the challenges in translating data to clinicians for practical use and must understand the impact of clinical care on outcomes. Ideally, epidemiologists and clinicians should form collaborative affiliations to study recurrent (and other) pregnancy complications to minimise the chance that relevant clinical influences are overlooked and that appropriate study designs and analyses are applied. Collaborations will also improve the likelihood that important information will be incorporated into clinical practice where appropriate. Importantly, clinicians in active practice should be encouraged to contribute to these efforts as a mechanism to help maintain clinical relevance. Perhaps it is time to establish an agenda for future recurrence research to bring together the appropriate participants and identify resources for examining this complex area. In this way, the field can move forward quickly, with focus, and produce answers to our patients' important questions of recurrence risk and prevention.

Ancillary