Long-term child follow-up after large obstetric randomised controlled trials for the evaluation of perinatal interventions: a systematic review of the literature
Miss MJ Teune, Department of Obstetrics and Gynaecology, Academic Medical Centre, PO Box 22700, 1105 AZ Amsterdam, the Netherlands. Email: email@example.com
Background Although the hope is that many perinatal interventions are performed with an ultimate aim to improve the long-term health and development of the child, long-term outcome is rarely used as a primary end-point in perinatal randomised controlled trials (RCTs).
Objective To evaluate how often and with which tools long-term follow-up is performed after large obstetric RCTs.
Search strategy We searched the Cochrane Library for Cochrane reviews published by the Cochrane Pregnancy and Childbirth Group for reviews on interventions that aimed to improve neonatal outcome.
Selection criteria Reviews on perinatal interventions that were not performed to improve the condition of the neonate were excluded. We limited our review to RCTs with more than 350 participating women. For each included study, we checked in Web of Science as to whether the researchers had reported on follow-up in subsequent publications.
Data collection and analysis Relevant information was extracted from these RCTs by two reviewers using a predefined data collection sheet. All information was analysed using SPSS 17.0 (SPSS Inc., Chicago, IL, USA).
Main results We studied 212 reviews including 1837 RCTs on perinatal interventions, 249 (14%) of which included 350 participants. Only 40 of 249 RCTs (16%) followed the children after discharge from the hospital to evaluate the effect of a specific perinatal intervention. The number of RCTs with long-term follow-up remained stable, with 10 of 67 RCTs (15%) reporting follow-up before 1990, 17 of 115 (15%) between 1990 and 2000, and 13 of 67 (19%) after 2000 (P = 0.68).
Conclusions Only a small minority of large perinatal RCTs report the long-term follow-up of the child. Future obstetric RCTs should consider performing long-term follow-up at the start of the trial.
The hope is that many perinatal interventions are performed to improve long-term child outcome. Nevertheless, most primary outcomes of randomised controlled trials (RCTs) are still short-term outcomes. We believe that continuous follow-up of children who take part in perinatal RCTs long after discharge from the hospital is necessary, because serious sequelae from perinatal complications frequently manifest themselves only after several years.1 However, long-term follow-up is time-consuming, expensive, often beyond obstetricians’ main research interests and falls outside the funding period of most perinatal RCTs. Consequently, the primary end-point tends to be selected from outcomes that occur before initial hospital discharge.2
Long-term follow-up is crucial, as perinatal interventions often aim to prevent morbidity in later life and to optimise neurodevelopmental outcome, thus changing the overall conclusion regarding the optimal diagnostic or treatment strategy in a pregnant woman. The administration of diethylstilbestrol during the 1950s or, more recently, the administration of antenatal thyrotropin-releasing hormone for the prevention of neonatal respiratory disease or antibiotics for spontaneous preterm labour are prime examples of interventions that turned out to have important negative long-term effects.3–10
The aim of this systematic review was to evaluate how often long-term follow-up is performed in large perinatal RCTs and whether or not this was planned before the start of the RCT. If follow-up was performed, we documented the percentage of children that were followed after discharge from the hospital and the follow-up methods that were used.
We believe that this information can provide guidance to future research teams performing clinical trials and will encourage the setting up of follow-up at the start of perinatal RCTs. In the case of limited funding or lack of infrastructure, it will help in the choice of follow-up tools, appropriate for the given circumstances.
We searched the Cochrane Library 2008 issue 4 for Cochrane reviews published by the Cochrane Pregnancy and Childbirth Group for reviews on interventions that aimed to improve neonatal and/or child outcome. We assumed that all important perinatal RCTs published would be included in one of the reviews of the Cochrane Pregnancy and Childbirth Group. All reviews on perinatal interventions that aimed to improve neonatal and/or child outcome were included. Reviews on interventions primarily looking at aspects other than neonatal and/or child outcome, such as labour progress, were also included when these interventions could affect neonatal and/or child outcome. Cochrane reviews of perinatal interventions that had no potential health benefit for the neonate or child were excluded (for example, the Cochrane review ‘Antibiotics regimens for endometritis after delivery’).
From the relevant Cochrane reviews, we selected studies reporting on RCTs with more than 350 women. Every included RCT was screened for statements on long-term child follow-up. RCTs without statements on long-term follow-up were cross-checked with the Web of Science to see whether long-term effects were reported in subsequent publications. We assumed that articles reporting on the long-term effects of a specific perinatal intervention would refer to the original article reporting this RCT.
Relevant information was extracted from these selected RCTs by two reviewers using a predefined data collection sheet. The following characteristics were extracted from the included RCTs: general study characteristics (country of investigation, year of publication, journal of publication, sample size, primary outcome if mentioned, effect of intervention, power calculation), whether long-term follow-up was performed, whether this follow-up was planned before the start of the RCT, and whether the overall verdict regarding the effect of an intervention was changed after long-term follow-up. If follow-up was performed, the duration of follow-up, follow-up rate and methods and instruments used were documented. All information was entered and analysed using SPSS 17.0 (SPSS Inc., Chicago, IL, USA).
Of the 304 reviews of the Cochrane Pregnancy and Childbirth Group, 212 contained potentially relevant information, as they reported on interventions to improve neonatal and/or child outcome. These relevant reviews included 1837 studies on obstetrical interventions, 249 (14%) of which reported on more than 350 women. Of these 249 studies, only 40 (16%) followed the children after discharge from the hospital to evaluate the effect of a specific perinatal intervention (Figure 1),3–95 and 209 studies performed no long-term follow-up (see Appendix S1 for references). Table 1 shows the characteristics of the included studies. Before 1990, follow-up was performed in 15% (10/67) of RCTs versus 15% (17/115) between 1990 and 2000, and 19% (13/67) after 2000 (P = 0.68). Large perinatal trials were most often performed in the USA (74/249), closely followed by the UK (47/249) and mainland Europe (48/249).
Table 1. Study characteristics
| Published |
|Before 1990||57 (85)||10 (15)|
|1990–2000||98 (85)||17 (15)|
|After 2000||54 (81)||13 (19)|
| Country of investigation |
|USA||61 (82)||13 (18)|
|UK||45 (96)||2 (4)|
|Mainland Europe||40 (85)||7 (15)|
|Canada||5 (71)||2 (29)|
|Australia/New Zealand||11 (65)||6 (35)|
|African country||17 (89)||2 (11)|
|Asian country||9 (75)||3 (25)|
|South America||11 (85)||2 (15)|
|Global studies||10 (77)||3 (23)|
RCTs with follow-up after discharge from hospital
Table S1 shows the characteristics of the follow-up studies. As mentioned above, 40 large RCTs performed follow-up of the child after discharge from the hospital. The duration of follow-up varied from 3 months to 40 years. Intentions for these long-term follow-up studies were mentioned in 21 of the 40 initial reports (53%).
Of the 249 RCTs, 145 reported a power calculation for the primary outcome in the initial report, but an adequate power calculation for long-term child outcome was reported in only six of the 40 RCTs (15%) in which long-term follow-up was studied. Not all liveborn infants included in the RCTs were approached for follow-up. The following reasons were often mentioned for not approaching all children: exclusion of children born in hospitals in specific countries or states; exclusion of children born before or after a specific date of birth; parents declining participation before the start of follow-up; inclusion of children born in hospitals with a follow-up rate of >80% only; inclusion of children with a high a priori probability of abnormalities only (for example, inclusion of children with abnormal neurological signs during the neonatal period); only a random sample of the study population was followed up after the neonatal period. Overall, the follow-up rate varied between 11% and 100%.
The effect of a perinatal intervention was judged to be positive in the short term in 18 of 40 RCTs, negative in two of 40 RCTs, no effect in 13 of 40 RCTs and unclear in seven of 40 RCTs. The overall conclusion with regard to the treatment effects changed in the majority of RCTs after long-term follow-up (24 of 40 RCTs). Of the 13 RCTs judged to have no effect after short-term analysis, six reported a different outcome after long-term follow-up of the child (46%).
Methods used for follow-up
In Table S2, the long-term outcomes and the methods used for follow-up are shown. The table shows that there was no overall consensus with regard to which methods and types of questionnaires or assessments should be used to measure specific long-term outcomes. In total, 28 studies evaluated long-term neurological development of the child. Of these, 28 studies,3,4,8,9,11,12,17,18,20,21,23–27,30–38,42–46,48,50–53,58,60–62,70–74,76–85,88–94 four used a questionnaire(s) only, eight used assessment(s) only, 15 used a combination of questionnaire(s) and assessment(s), and one performed a neurological examination only.
We found that, of 249 large RCTs on perinatal interventions, only 40 (16%) followed the children after discharge from the hospital to evaluate the effect of a specific perinatal intervention. Intentions for these long-term follow-up studies were mentioned in 21 of the 40 initial reports. Although there is an increasing awareness of the importance of long-term follow-up after perinatal RCTs, we found no clear trend indicating more long-term follow-up in studies over time.
This review also shows that there is no consistency in the methods used for the follow-up of children. Some differences may be justifiable. When, for example, one is evaluating the effect of an intervention, such as zinc supplementation, on growth, one is particularly interested in differences in anthropometric measurements later in life and the neurological development of a child may not be the primary concern. Nevertheless, the studies evaluating neurological development used very different methods with assessments performed at very different ages. This finding indicates the need to improve standardisation and comparability of methodology and data collection for long-term outcomes in perinatal RCTs, so that future RCTs with comparable research questions will have the same primary outcome and will use the same long-term follow-up tools.
A relative limitation of this review could be the search strategy used. We decided to search for relevant RCTs on long-term outcome in the Cochrane Library. The Medline search did not provide us with sufficient relevant RCTs because of a lack of good MeSH terms. Therefore, it is possible that some large perinatal RCTs with follow-up are missing in this review because these trials were not included in reviews published by the Cochrane Pregnancy and Childbirth Group.
Another relative limitation is that we restricted our search to RCTs with >350 women. The cut-off point of 350 women was selected because large RCTs have the power to detect significant differences in short-term as well as long-term outcomes. As a consequence of this cut-off point, smaller RCTs with long-term follow-up were not included in our analysis. Therefore, our results may not be applicable to smaller sized RCTs. Future reviews comparable with ours are needed to determine whether different sample sizes lead to different results.
Follow-up studies aimed at determining child outcomes must be designed carefully. The choice of the appropriate child outcome measure should be in line with the original study question. Following from our results, the outcome measures are most often related to brain development, or concern neuromotor or developmental measures. In addition, child growth measures, child health and metabolic outcomes may be relevant to the study question.
For all of these outcome measures, the development of prediction models for long-term child outcomes based on short-term neonatal outcomes might be an alternative to expensive and often underpowered follow-up studies.96–98 To explore the validity of these models, a first step would be to study validity in reports with long-term follow-up. Modelling has several advantages. It is comparatively inexpensive and free of ethical concerns over consenting issues. It is also fast: a computer model can simulate in minutes, whereas follow-up lasts years. Of course, modelling also has limitations. Failings in model theory or logic, inaccuracies in model parameters and the omission of key factors can all invalidate the results.99
Nevertheless, in the majority of studies, (real-time) follow-up is inevitable to evaluate the effect of the intervention on the health of the child. This follow-up can be performed through questionnaires and/or assessments. Questionnaires, such as the Ages-and-Stages Questionnaire (ASQ), have the advantage that they are relatively inexpensive and easy to organise. A disadvantage is that they are designed as developmental screening tools and therefore are less useful for the detection of mild problems that tend to be common in preterm-born children.98 Other disadvantages are that parents frequently over- and under-report the health status of their child, and response rates vary considerably between studies.99,100
Although questionnaires, such as ASQ, are designed to identify severe developmental delay, the Bayley Scales of Infant Development (BSID) is the most widely used test to assess infant and toddler development in the areas of cognition, language, motor development and behaviour for those aged 3 years and under.1,101 Developmental scores range from severe delay to accelerated development, and comparisons of mean scores per group can therefore show whether one treatment is superior to another. However, at this young age, mental and motor development are still intermingled, and delays may not reflect impaired development. Moreover, at the age of 5 years and beyond, disabilities become apparent in the different developmental domains, and therefore, at this age, developmental outcomes have a greater predictive value for later in life.102,103
It would be a step forward if RCTs around the world used a similar follow-up protocol. The choices to be made would then be whether to use a questionnaire and, if so, with what aims. In addition, the age at which an assessment of child development is carried out needs to be discussed. Choices need to be made, especially for situations with restricted funding. Is an assessment before 3 years preferable, despite the fact that it is still a global measure, or should development at a later age be used to indicate whether there are disabilities in different developmental areas.98,104–106 When it is to be expected that certain obstetrical treatments will have effects on certain brain areas, it is only logical to assess child functions related to these areas.
Only a small minority of large perinatal RCTs report the long-term follow-up of the child. Future obstetric RCTs should consider performing long-term follow-up at the start of the trial. We recommend the development of standard guidelines for follow-up after perinatal RCTs, and encourage the setting up of follow-up at the start of perinatal RCTs. In the case of limited funding or a lack of infrastructure, the choice of follow-up tools appropriate for these circumstances should be made.
Disclosure of Interests
ZA is joint Co-ordinating Editor for the Cochrane Pregnancy and Childbirth Group.
Contribution to authorship
MJT is the principal contributor to the study described in this article. BCO, BWJM and AGvW developed the initial study protocol. BCO, BMJW, AGvW and MJT participated in the study design and coordination. MJT wrote the first draft of the manuscript. All other authors commented on this draft and contributed to the final manuscript.
Details of ethics approval
No ethics approval was required.
This study was financed by grant 80-82325-98-9010 from ZonMW, The Netherlands Organisation for Health Research and Development, The Hague, the Netherlands.