Portal de Periódicos da CAPES

Fetal monitoring in labor: Implications of evidence generated by new systematic review

2015; Informa; Volume: 95; Issue: 1 Linguagem: Inglês

10.1111/aogs.12830

1600-0412

Amar Bhidé, Edwin Chandraharan, Ganesh Acharya,

Neonatal Respiratory Health Research

ST waveform analysis of fetal electrocardiogram (STAN) was introduced in clinical practice after evaluating its potential to diagnose fetal hyoxemia/acidosis in several experimental animal studies and human clinical trials in the 1980s. Many randomized controlled trials (RCTs) have since been performed, and the most recent one was published in August 2015 1. This issue of the AOGS contains a systematic review with a meta-analysis of six RCTs (26446 women) published so far 2. The total number of included women is impressive. However, the controversy surrounding the superiority of electronic fetal heart rate monitoring using cardiotocography (CTG) plus STAN, over CTG alone for monitoring fetal condition during labor, continues. Use of STAN as an adjunct to CTG is claimed to have several advantages over CTG alone. These include a reduction in the rates of fetal metabolic acidosis, operative delivery for fetal distress, and need for fetal blood sampling and improved fetal and neonatal outcomes (death or serious morbidity). The present systematic review has confirmed a lower rate of metabolic acidosis, a reduction in the rate of fetal scalp blood sampling (FBS), and a reduction in the overall rate of operative vaginal delivery. There was no reduction in the rate of operative delivery for fetal distress, and the fetal and neonatal outcomes were no different. The authors of the systematic review conclude that the overall effects of CTG plus STAN were minor, and question the place of CTG plus STAN in contemporary obstetrics. Does this mean that units currently using STAN should stop using it and should go back to using CTG alone with or without FBS back-up, or should units currently not using STAN consider its introduction based on the reported reduction in metabolic acidosis, rate of FBS and overall operative vaginal delivery rate? The authors of the systematic review have argued that appropriate use of surrogate endpoints, such as metabolic acidosis, requires a direct correlation between the surrogate and the truly important outcome. They discourage excessive emphasis on the positive results for metabolic acidosis, and suggest the need for caution in interpreting the statistically significant finding of reduced rate of metabolic acidosis 2. Others may disagree, as metabolic acidosis is generally considered a relevant measure of fetal wellbeing in clinical practice. The Belfort study 1 used death or composite morbidity as the primary outcome criterion. The estimated prevalence of this outcome was 1.75%, and the trial was powered on an estimated 40% reduction in the primary outcome. In fact, the prevalence of the primary outcome was 0.72% in the masked group. A 40% reduction is a reasonable expectation, but a more modest reduction can still be clinically meaningful. Sample size increases exponentially as the effect size is reduced, and this can be clearly seen from Table 1. Therefore, although this systematic review has failed to show significant reduction in hard outcomes such as death, neonatal seizures and neonatal encephalopathy, it has not shown that there is no reduction in these outcomes. Of course, the sample size required to confirm no effect of STAN on these clearly important outcomes is so large that it is debatable whether this number will ever be reached by a systematic review, let alone by an RCT. We believe that a reduction in fetal blood sampling rate, operative vaginal delivery rate and metabolic acidosis are relevant, because they are the few outcomes for which the systematic review is powered adequately. However, because hard outcomes (death, encephalopathy, seizures) are unchanged, there is a risk of the exercise of meta-analysis degenerating into ‘insufficient evidence about everything…’. 0.8%4 2.8%5, 0.39%6 2.4%7 0.6%4 12.9%8 21%4 3.1%4 5.0%10 No intrapartum fetal blood sampling seems to have been performed in the Belfort trial. The criteria for intervention also appear to have been modified in RCT. The STAN guidelines used in clinical practice in most countries recommend intervention after three biphasic STAN events with intermediary CTG, whereas two events are sufficient for intervention with an abnormal CTG. STAN guidelines also stipulate intervention only if magnitude of the observed (i.e. episodic or baseline) T/QRS events reach the recommended threshold. In the Belfort trial, these recommended guidelines were simplified. CTGs were categorized into green, yellow and red zones. Intervention was recommended for any ST event in the yellow zone CTG regardless of the type, magnitude or the number of ST events. Expedited delivery was recommended for red zone regardless of the type, magnitude or the number of ST events. This approach was different to the standard STAN guidelines that were developed based on research evidence and that are used in Europe, Asia and Australia. From the published material it is not clear what proportion of CTGs were classified in the three different zones in each arm of the trial. If intervention was primarily based on ST events in only a small proportion of cases, one is unlikely to see a difference between the two arms. We contacted the lead author of the RCT for clarification (Dr. Belfort, personal communication). He explained that the modification was recommended by the manufacturers of STAN system (Neoventa). Moreover, he no longer has access to the data, and the data are with the US National Institute of Health (NIH), the funder/sponsor of the study. He forwarded our query to NIH (November 10th and 16th, 2015), but we have not heard from them yet. It is our opinion that this information would be of great importance in order to judge the added value of the latest RCT published in the New England Journal of Medicine 1. There were over 26000 women included in the six RCTs published so far. A risk reduction even by one fourth (0.25) would be clinically useful for serious conditions such as fetal or neonatal death. However, even for a 50% reduction, a sample size of over 150,000 is needed to show superiority of the intervention, given the low prevalence of death (Table 1). What is evident from Table 1 is that the observed prevalence of adverse events in the RCTs is lower than reported prevalence in everyday practice. The contribution of Hawthorne effect is well known. What is unclear is if this reduction is solely attributed to Hawthorne effect or if it is also related to improved understanding associated with training/teaching in CTG interpretation. It has been known for some time that outcomes are better just through participation in a study 11. In fact, Individual Participant Data (IPD) meta-analysis of RCTs on CTG plus STAN versus CTG have shown a significant improvement in the outcomes during the second half of the trial, demonstrating the impact of training and education 12. Such an impact of training on clinical outcomes has been reported from an observational study from St George's Hospital in London 13. One of the authors of this editorial has previously debated about the futility of intrapartum FBS, and, with others, called for interpretation of CTG based on the understanding of fetal physiology rather than relying on FBS 14-16. Regular training in CTG and STAN would be a mandatory part of any clinical trial protocol evaluating the performance of CTG/STAN trials. Therefore, it is easily possible that the observed lower prevalence of adverse outcome is at least partly related to the regular training improving interpretation of electronic fetal heart rate monitoring. The overall prevalence of FBS in the systematic review was 13.7% in both CTG and STAN arms combined 2. Although reduction in the use of FBS in the STAN arm is statistically significant, it was still used in 10.2% of cases in the STAN arm. We argue that this is only valid in the setting of RCTs, but not in actual practice. In institutions using STAN technology, intrapartum fetal blood sampling is uncommonly performed in an everyday clinical setting. The use of fetal scalp sampling at the authors’ hospitals has been virtually eliminated after the introduction of STAN in 2002. At St. George's Hospital London, UK, the FBS rate was 35 per year (0.7% of all births) in 2005, which reduced to 13 (0.2% of all births) in 2007, seven (0.1% of all births) in 2010 and to only three (0.05% of all births) in 2012. Between 2012 and 2015 (with approximately 15,000 births during this time), only two FBSs were performed (unpublished data). In the University Hospital of Northern Norway, Tromsø, Norway, fetal blood sampling has been abandoned after the introduction of STAN in routine clinical practice. The SR has shown that clinical outcomes with or without using STAN are comparable, and benefits or harms are likely to be marginal. In such cases, the decision to use STAN should rest on logistics or economical considerations. Cost effectiveness analysis of CTG plus STAN has been published previously 17. In that study, the cost was calculated based on trial data, and FBS was costed for 11% of women in the CTG plus STAN arm. Given that this frequency is not seen in clinical practice, the reduction in FBS should be costed on the frequency seen in everyday practice. Conversely, it can be argued that a reduction in metabolic acidosis is irrelevant as long as the rates of intrapartum fetal death, neonatal death or serious neonatal morbidity are unchanged. Therefore, the cost analysis should include incremental costs of the technique (electrodes and STAN machine as opposed to conventional CTG machine) balanced with savings by reduction in both FBS use and operative vaginal delivery (all indications). We have not seen any such analysis published so far, and the jury is still out. Publication of this systematic review and meta-analysis further advances our knowledge about the value of STAN technology for intrapartum monitoring. Outcomes that are directly associated with fetal/neonatal death or neonatal neurodevelopmental deficit and long-term disability are so uncommon that additional trial data are unlikely to result in statistically significant results. Surprisingly, FBS is still being used in clinical practice without any RCTs to confirm its safety and efficacy in improving any of the above hard outcomes, despite a recent systematic review 18 suggesting that FBS may actually increase the rate of operative deliveries. Conditions of the RCTs do not map well to everyday clinical practice. Hard clinical outcomes with or without the use of STAN are comparable, although metabolic acidosis rate is significantly lower amongst babies monitored by STAN. Therefore, based on clinical data on hard outcomes, it appears that there is insufficient evidence to introduce STAN in routine practice. However, the decision as to whether one should continue to use or abandon the use of STAN in the units where it has already been a routine method used for monitoring fetal heart rate in labor must be based on observed clinical outcomes, logistics, patient safety issues and economic considerations. Similarly, maternity units contemplating introducing STAN for intrapartum fetal monitoring should consider whether significant reductions in overall metabolic acidosis rate, FBS rate and total operative vaginal delivery rate represent tangible clinical or economic benefits. In either case, ongoing education/training of doctors and midwives, along with evaluation of their competency in interpretation of electronic fetal heart rate monitoring methods, should continue to remain in focus.