Choosing outcomes in pregnancy research

When asked to write an editorial on my personal experiences with pregnancy research, I accepted with enthusiasm but then found the task to be quite difficult. The topic is so broad that I struggled with the approach. Which of so many aspects should I touch on? There is the mastery of: the relevant published literature in pregnancy, the current relevant standard of care which must involve multidisciplinary collaboration, issues in internal medicine outside pregnancy as they apply to the relevant medical condition in pregnancy, choosing outcomes of relevance for both the mother and baby(ies), and then mobilization of a local team and international colleagues. Upon reflection, I decided to focus my comments on my struggles with outcomes that cover maternal, fetal and neonatal issues. Fetal and neonatal outcomes must also take into account the fact that women may deliver over a range of gestational ages. This will be true of any condition, pre-existing or gestational, that requires treatments for weeks or months before delivery. Important outcomes for the fetus before 20 weeks (e.g. miscarriage) are different from those of the very preterm baby (such as bronchopulmonary dysplasia). Chronic lung disease is not even possible for the baby born after 32 weeks’ gestation, and by definition, hypoxic–ischaemic encephalopathy and its devastating neurological implications can occur only at term. The absolute incidence of these adverse outcomes, however, is usually very low, even when the maternal condition in question (such as renal disease) is associated with substantially increased relative risks in pregnancy. As such, it is commonplace in perinatal research to study composite outcomes. This practice has been driven by the fact that outcomes of particular interest are devastating (e.g. maternal death) but not common enough to make feasible studies that examine these outcomes individually as their primary or secondary outcome. Combining these outcomes increases the anticipated incidence of the primary outcome and brings down the sample size. An example of this is the Pre-eclampsia Integrated Estimate of RiSk Score (PIERS), an outcome prediction model for adverse maternal outcomes in women admitted to hospital with preeclampsia. A Delphi consensus process was used to derive a composite outcome that covered concerns specifically related to the maternal central nervous, cardiorespiratory, haematological, hepatic and renal systems. 1 In a randomized controlled trial of treatment of gestational diabetes (versus standard care), the primary outcome was a composite of perinatal death, shoulder dystocia, bone fracture and/or nerve palsy. 2 Although combining multiple uncommon outcomes into a composite outcome is appealing from the perspective of sample size and feasibility, there are some serious concerns that arise. First, are the component outcomes of equal importance? The answer is ‘yes’ for a composite of neonatal morbidity associated with preterm birth, such a one or more of retinopathy of prematurity, necrotizing enterocolitis, bronchopulmonary dysplasia and/or severe brain injury. All of these morbidities are serious in the short-term and all have potential long-term implications. Clinicians and parents would wish to avoid each and every one of these outcomes. However, if component outcomes are not considered to be of equal importance, the results of a study may be driven by a less important but more common component. An example is the HYPITAT trial of induction of labour versus temporizing care for women who are at term and who have either gestational hypertension or non-severe preeclampsia. 3 The primary outcome was a composite of poor maternal outcome, defined as maternal mortality, maternal morbidity (eclampsia, HELLP syndrome, pulmonary oedema, thromboembolic disease and placental abruption), progression to severe hypertension or proteinuria and major postpartum haemorrhage (.1000 mL blood loss). Although induction of labour was associated with a lower incidence of the primary outcome (31% versus 44% among women allocated to expectant management), the results were driven by the between-group difference in the incidence of severe hypertension. Antihypertensives were used to treat only severe hypertension. Would induction of labour still be beneficial in settings where antihypertensives are used to treat non-severe hypertension? Second, could the study intervention cause the component outcomes within a composite to move in opposite directions? If so, the composite must be considered to be fatally flawed. This is the primary reason for not combining maternal and perinatal outcomes. For preeclampsia at ,34 weeks, delivery is always the best approach for the mother whose exposure to the preeclampsia process and its risks is therefore minimized, but delivery at early gestational ages may be associated with substantial neonatal mortality and morbidity.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.