Results

Of the 334 babies randomised, 66 (19.8%) died before discharge from hospital. Fifty nine children (17.7% overall or 22.0% of those discharged alive) were considered to be lost to follow up. The loss was very similar in the two trial arms (fig1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Derivation of children in the 2 year follow up.

Five children (1.5%) were known to have died between discharge and the 2 year follow up. The causes of the three deaths in the etamsylate group were given as extensive severe bronchopulmonary dysplasia, sudden death (microcephaly), and possible sudden infant death syndrome. The two deaths in the control group were recorded as sudden death (regurgitation and pulmonary dysplasia, with spastic tetraplegy) and impaction of bowel with strangulated hernia.

Information was available for 204 surviving children at the 2 year follow up. The median age at assessment for both groups was 106 weeks (IQR 103–113). For five children (two receiving etamsylate, three controls), no follow up information was available from the local paediatrician, but was ascertained from the parental questionnaire.

Table 1 gives information about the characteristics of the babies at trial entry and their use of etamsylate, for all those randomised and for those assessed. Those who died were more likely than the survivors to have been of lower birth weight and shorter gestational age, to have had cerebral problems on ultrasound, and to have not completed the full 16 dose course of etamsylate. The children lost to follow up did not differ significantly between the two trial arms, and seemed more similar to those assessed at 2 years of age than to those who died (data available but not shown).

Discussion

Etamsylate prevents capillary bleeding during surgery for adults, and therefore may reduce PIVH in preterm neonates. Horbar7has summarised possible mechanisms of action in terms of stabilisation of capillary membranes and an increase in platelet adhesiveness. It may also have direct effects on the cerebral vasculature by inhibiting prostaglandin synthesis, and indirect effects by closing the patent ductus arteriosus, as suggested in a small trial by Amatoet al.8

The importance of PIVH is as a predictor of longer term outcomes. In this study, 31% of the 268 babies who had survived to the time of discharge from hospital had been diagnosed with PIVH (with the remainder either known not to have such haemorrhaging or whose haemorrhage status was not known). The status at 2 years of age of 74 of the 268 was not known. On the basis of the remaining 194 children, a diagnosis of PIVH did not appear to significantly raise the risk of the primary adverse outcomes used in the trial. The risk of death or severe impairment or disability was 13.3% (8/60) in the PIVH group compared with 10.4% (14/134) in those without such a diagnosis (RR 1.2; 95% CI 0.66 to 2.19). Similarly, the risks of death or any impairment or disability were 30% (18/60) compared with 23.1% (31/134) (RR 1.27; 95% CI 0.81 to 1.98). Given the imperfect predictive value of PIVH as the main outcome for this trial, the follow up reported here clearly adds useful information about longer term implications of interventions in the neonatal period. This paper is currently the only published follow up from a trial of neonatal etamsylate. Provisional findings from the four year follow up from the UK multicentre trial were reported in abstract9 but not published in full.

A third of the children in the EC trial died or were classified as having an impairment or disability at the age of 2 years. There is little evidence from this trial, however, that a policy of neonatal etamsylate would reduce the burden on their parents or on the health services.

There are a number of possible reasons for this finding. It could be that more children in the control group who were lost to follow up suffered an adverse outcome. This seems unlikely given the similarity in loss to follow up between the randomised groups.

It could be that a potential lack of blinding to allocation in Greece led to a bias in the paediatrician's assessment of the child's status. Again, this seems unlikely, partly because such bias was more likely to be in favour of etamsylate (the experimental treatment), and partly because the findings in France and Greece are similar.

Another possibility could be that the beneficial effects of etamsylate were diluted because it was not given soon enough after birth. This hypothesis is not supported by the analysis stratified by age at randomisation.

Perhaps the trial population, by including babies who, at 31 and 32 weeks gestation, had a relatively high chance of a good outcome, was not at sufficiently high risk. This seems unlikely to be a reason because about a third of the babies had a bad outcome, and the analysis stratified by gestational age did not suggest a greater effect in the lower gestational age group.

Alternatively, perhaps the population was at too high a risk, given that nearly a third already had cerebral problems on scans at the time of trial entry. Analysis of the primary outcomes stratified by known cerebral problems were, however, unable to detect a significant benefit of etamsylate.

It may be that, had the assessment been performed by a research developmental paediatrician in the child's home rather than in a normal follow up clinic, more subtle effects may have been detected. However, this would apply equally to babies in both trial arms.

Although this is the largest trial of neonatal etamsylate, it may still not be powerful enough to detect uncommon effects. The 95% confidence intervals are still wide and so can only exclude the possibility of a benefit if it was less than about 20–25%.

The most likely reason for not being able to detect a sizeable benefit of the etamsylate policy in this trial is that such a benefit does not exist. Alternative strategies such as neonatal indomethacin10 and delayed umbilical cord clamping11 still need to be evaluated in well designed studies.