These 3 articles have just been published, all show no difference in long term outcomes between the randomized groups. What does that mean for the impact on therapeutic decision-making?
The first of them, Natalucci et al is a longer-term follow up of a trial that already reported their 2-year primary outcomes. It was an RCT of infants of 26 weeks gestation or more and less than 32 weeks, who received either 3000 iu of erythropoietin at < 3 hours of age and then 2 other doses at 12 to 18 hours and 36 to 42 hours. There was no impact on the Bayley MDI at 2 years, and now they have shown in the 77% of children left in the trial (total n= 345) that there was no difference between groups on a test of cognition, or in cerebral palsy, disabling CP, or hearing or visual problems. This study complements the PENUT trial which included babies down to 24 weeks gestation (up to < 28 weeks) and gave a dose of 1000 iu/kg every 48 hours for 6 doses and then 400 iu/kg 3 times a week until 33 weeks. That study also showed no benefit on long term outcomes.
The next study, Roze et al TRIOCAPI , randomized babies born between 24 and <28 weeks who did not have an IVH on initial head ultrasound. They had a screening cardiac ultrasound at 12 to 24 hours of age and, if the ductus was “large” (calculated as > 2.26 – (0.078 x postnatal age in hours) mm), they received either ibuprofen or placebo. The primary outcome was survival without cerebral palsy at 2 years of age; among the 228 babies randomized in the 2 groups, this outcome was not different by treatment group, at just over 71%. This was double what they were expecting in the control group, as outcomes have improved dramatically in France for these very immature babies, so they were somewhat underpowered. They did show a decrease in pulmonary haemorrhages in the first 3 days (those requiring an increase in FiO2 > 20% or an increase in mean airway pressure > 2cmH2O) from 8% to just under 2% with ibuprofen; at 3 days of age, they were much more likely to have closed the PDA (66% vs 17%). There were a large proportion of the babies, 62% of controls and 17% of the ibuprofen group, who received open-label ibuprofen after the first 3 days.
The third of these studies Tauzin et al, PRETTINEO is the first, I think, controlled trial of premedication for neonatal intubation that has published long term follow up and has a sample size large enough to have reasonable power. The drug regimes compared for non-emergency intubations in the NICU are mentioned in the title. Atropine at 15 microg1kg was given to everyone, followed by either propofol (2.5 mg/kg for babies >1kg, and 1 mg/kg <1kg) or atracurium 0.3 mg/kg and sufentanil (0.2 microg/kg >1kg and 0.1 microg/kg <1kg). The initial publication of the acute results showed about the same incidence of the primary outcome, prolonged desaturation, in the two groups. The propofol group required many more extra doses of medication to achieve “adequate anaesthesia”; they were also less well sedated and required a longer time to be intubated. However, they started to breathe again more quickly, the atracurium sufentanil group taking a median of 33 minutes compared to 14 with propofol. The follow-up data are mostly from the Ages and Stages Questionnaire, and showed no difference between the groups. However, of the 166 babies included in the follow-up only 118 had data from the ASQ included, the others haveing their data imputed, many more in the propofol group (40%) than the atracurium/sufentanil group (19%). With this limitation in mind, all of the ASQ scores were basically identical between groups.
What should we do when we have results like these 3 trials, showing no difference in the long term outcomes or mortality between two treatment approaches? I don’t thin kit necessarily means you should throw the treatment out of the door, if there are really no differences in long term outcomes between 2 treatments then there are a few questions we should ask:
How reliable are the results?
What differences between groups are compatible with the results?
How applicable is this to my practice?
What are the short term impacts of the treatment?
What are parents likely to prefer?
How reliable are the results? Is the trial likely to be unbiased? Or are there potential sources of bias in the trial design or reporting of the results? This is a huge subject, but some things can improve your confidence that there really is no difference between the groups, such as a trial with masked allocation (and masked intervention, if possible) funded by an independent source that was registered before the trial started and reports the same primary (and a limited number of secondary) outcomes as are in the registration documents.
What differences between groups are compatible with the results? A trial may be called a null trial because the results did not reach a threshold of “p<0.05” but the confidence limits of the trial should be examined, a small to moderate size trial may still be compatible with a large, clinically important, difference in the treatments.
How applicable are these results to my practice? If the comparison group is not managed as they are managed in my practice, if the eligibility criteria eliminate many babies that I look after, or if control group outcomes are dramatically different to my patients, then the applicability to my practice may be very limited.
What are the short term impacts? We have become so focused on long term outcomes, very often survival without disability, that the impacts of short term outcomes may get lost. So a treatment that doesn’t impact survival or long term outcomes might well have advantages that are worthwhile. Such as reducing severe IVH despite no change in developmental progress or reducing the need for retinopathy therapy despite no final impact on blindness. If those short term benefits are achieved without short term adverse effects, that might be an indication for using a treatment.
What are parents likely to prefer? Outcomes which are important to parents should be a major part of our considerations, even if long term benefits are few or unproven. To return to the example above, would a parent prefer that their baby does not have a severe IVH, even if that doesn’t necessarily improve their long term outcome? If so, then prophylactic indomethacin should be considered, especially as the large, high-quality trials (Ment Indo IVH Prevention Trial and TIPP) showed no substantial difference in adverse events.
To apply these question to our three new null publications.
How reliable are the results?
For the trial of Natalucci, I would say that I can’t find any important potential bias in this trial design, a multicentre double-masked trial with the primary outcome as initially specified.
The TRIOCAPI trial is also well done, again a multicentre masked RCT with a pre-specified primary outcome.
For the PRETTINEO trial, the low rate of follow up and the reliance on the imputed outcomes makes me rather hesitant, especially with the very low rate of follow up in the propofol group.
What differences between groups are compatible with the results? This is, of course, a consideration of power and of confidence intervals.
For the Natalucci trial the primary cognitive score was almost identical in the 2 groups, and the results were compatible with a true difference between groups of a 3 point decrease in scores to a 2 point increase (approx), and therefore very little chance of a significant adverse impact of Erythropoietin.
The results of TRIOCAPI showed, again, almost identical rates of the primary outcome, but, as a smaller trial, the results are compatible with a 17% relative reduction or a 16% increase in the outcome of survival without cerebral palsy. The absolute risk difference in CP between groups was about 5% and the 95% limits of the absolute difference are between about a 9% lower frequency with indomethacin and a 6% higher frequency. The differences in disabling CP (GMFCS >2) were tiny (3 patients in the placebo and 2 in the indomethacin groups).
For PRETTINEO, the confidence intervals for the difference in “survival without neurodevelopmental delay” are very different if you include the imputed values or not, and therefore I would say that you can’t really rely on the confidence intervals.
How applicable is this to my practice?
The Natalucci trial is relatively applicable, but it excluded the highest risk babies of under 26 weeks. Survival and other outcomes among enrolled babies are not very dissimilar to mine.
TRIOCAPI again excluded the most immature babies <24 weeks, but included 24 week infants (French NICUs have a higher rate of comfort care in the delivery room at 24 weeks than we do, so the proportion of enrolled 24 week infants is somewhat lower), the main thing that makes me concerned about applicability is the very high rate of early PDA treatment among placebo group babies, which is much more, I guess, than a similar group of babies in my NICU. But, whether that would have an impact on the rate of CP or other developmental delays, I doubt.
PRETTINEO compared propofol premedication to a regime that is somewhat similar to what I use, which is atropine/succinylcholine/fentanyl. Succinylcholine has a much shorter duration of action than atracurium, so the babies start breathing gain much faster (sometimes it is too fast and we have to give a second dose), so I would say somewhat relevant.
What are the short term impacts?
Natalucci’s trial showed no adverse effects; the PENUT trial also showed no evidence of any adverse effect, the previous concern about a possible increase in retinopathy with some regimes is not born out by this new data. I can’t find a report of transfusion requirements in the Nataluci trial publications, but PENUT showed that the proportion of babies who never needed a transfusion increased from 13% to 28% with their erythropoietin regime and the median number of transfusions decreased from 4 to 2.
TRIOCAPI also showed little in the way of adverse impacts of early targeted ibuprofen treatment. There were a lot fewer pulmonary haemorrhages 2% vs 8% in the first 3 days of life, but somewhat more GI perforations, 8.8% vs 3.5%. The Cochrane review of prophylactic ibuprofen (not exactly the same I know, but ibuprofen given in very early life, also shows more GI perforations compared to no treatment, but is only based on 2 small trials with a total of 167randomized babies and 14 events, the confidence intervals are very wide and include no difference. GI bleeding is also more common in the Cochrane review, although that is supposed to be “statistically significant” it is largely based on 2 very small Thai trials that had enormous rates of GI bleeding.
PRETTINEO in their initial publication showed a high frequency of hypotension after propofol, which received treatment on 2 occasions, No other major difference in short term impacts was shown, the proportion desaturating during intubation was similar. There were many more intubations on the first attempt in the atracurium/sufentanil group.
What does mean for the clinical implications of these null studies?
Early Erythropoietin, doesn’t seem to offer any neuroprotection when you put it into the context of other trials. It does seem to reduce the proportion of babies who receive a transfusion and the total volume of blood transfused, without any adverse effects. I think that is an impact that might be valuable to some parents, and that therefore it would be reasonable to offer it as an option to parents with babies at risk of being transfused. If we couple it with delayed cord clamping and greater efforts to reduce blood losses, by taking initial blood work from the blood left in the placenta for example, we could probably do even better to reduce transfusions, which I know are extremely safe these days, but still not as safe as not having a transfusion!
Early targeted ibuprofen was something that we introduced into our unit as a pilot after the Kluckow trial, partly because we had been through a phase of having quite a lot of pulmonary haemorrhages. This trial means, I think that we should rethink that approach. It confirms that there are fewer pulmonary haemorrhages leading to respiratory deterioration with this approach, but there may be an increase in GI perforation. Is there a better way to target those at risk of pulmonary haemorrhage to change the risk:benefit ratio?
The PRETTINEO trial confirms to me that propofol is not a good idea, even though there was no adverse long term impact (with the limitations already mentioned) intubation took longer and more attempts and there was more hypotension. Maintaining the use of a regime with a short acting muscle relaxant seems to me optimal, as long as a potent analgesic is also used.
Neonatal Research 
Thanks Keith! Great post on how to critically assess the evidence! I have just sent it to our staff and fellows. All the best