Why did the authors perform a meta-analysis of studies with primary endpoints they consider clinically unimportant?

Abstract

The meta-analysis of Blix et al. on ST waveform analysis (STAN) 1 confirmed the results of a previous meta-analysis 2. Their conclusions, however, are based on presuppositions that are highly questionable. The systematic review by Blix et al. 1 was prompted by a recently published American randomized controlled trial (RCT) on STAN 3. However, this RCT was performed on a predominantly low-risk population with guidelines for CTG interpretation and threshold for STAN intervention different from those applied in all European RCTs and in clinical use. Furthermore, the experience of the health workers with STAN in this trial must have been minimal beyond the pre-trial courses. The only aim of intrapartum fetal monitoring is to diagnose hypoxia as being a major preventable risk factor of neonatal encephalopathy (NE) and future cerebral dysfunction. Metabolic acidosis is a quantifiable measure of severe intrapartum hypoxia and is associated with perinatal death, NE and cerebral palsy 4. That is why most of the RCTs on STAN used metabolic acidosis as their primary outcome. The international consensus statement includes metabolic acidosis as a prerequisite for a causal relation between cerebral palsy and intrapartum events 5. Metabolic acidosis at birth is also regarded an essential criteria for neonatal cooling therapy in case of severe birth asphyxia at term. The evolving evidence on potential long-term effects of apparently minor insults in the perinatal period call for cautiousness regarding metabolic acidosis in case of a vigorous neonate as a trifle, based on clinical data from a single study 6. Blix et al. showed that the use of STAN was accompanied by a significant 36% reduction in cord metabolic acidosis at delivery 1. The authors disqualify metabolic acidosis at delivery as a surrogate endpoint 1 and argue that the majority of neonates with metabolic acidosis show a normal neurodevelopment and cognition at the age of six years 6. They missed that 10.3% of the cases with metabolic acidosis either had died due to perinatal asphyxia or had a severe neurodevelopmental disorder 6. If, according to the authors' own calculation, 401 women need to be monitored with STAN to avoid one case of metabolic acidosis 1, every 10th of those prevented cases (1/4010, Norway: 15/year) would otherwise have a severe adverse outcome (disability or death). Do the authors regard these 15 preventable severe outcomes as not relevant? The authors also showed that the use of STAN led to a significant 8% reduction of operative vaginal delivery 1. Surprisingly, the authors conclude that this finding has no practical significance. In a Norwegian national perspective (based on data for 2013), 493 vaginal operative deliveries annually are preventable by STAN monitoring. Do the authors really regard that reduction as not important? The authors state: “Absolute effects of STAN were minor” 1. If the numbers given above are to be considered unimportant, further realistic efforts to reduce risks and complications in modern obstetrics seem mostly unwarranted. The authors conclude: “There is not enough evidence to justify the use of STAN in contemporary obstetrics” 1. We assume the authors recommend a continued use of CTG, with or without fetal blood sampling, for intrapartum fetal monitoring and we look forward to have an updated justification for the use of these methods.