Statistics from Altmetric.com
A 25 week gestation male infant, birth weight 520 g, is transferred ex utero to your neonatal unit for intensive care. On day 19 he remains ventilator dependent and is hypoxic on 60–95% oxygen. The oxygenation index (OI; a measure of respiratory failure) is 18. Chest x ray shows clear lung fields. Echocardiography by a senior paediatric cardiologist shows evidence of pulmonary hypertension. In view of these findings it is felt that a pulmonary vasodilator may help. You discuss entering the INNOVO trial (a multicentre randomised controlled trial of addition of inhaled nitric oxide (inNO) to babies with severe respiratory failure) with the parents, who agree; the baby is entered into the control (no addition of inNO) arm. In spite of this, parents ask for “everything” (including inNO) to be tried. Reluctantly you agree and inNO is administered as per the trial protocol (you inform the trial coordinators). The baby does not improve, and dies 24 hours later. Was it reasonable to administer inNO to this baby?
Numbers needed to treat
The “NNT” (number needed to treat) is seen as the hallmark “statistic” in evidence based medicine. But what is it, and how should it be interpreted?
The NNT is “the number of patients needed to treat with A (compared with B) to prevent one defined outcome in a specified time period”. It is the inverse of the risk difference between the two treatments. (For example, if 5% of children with croup who receive steroids are admitted, and 25% of those who receive placebo are admitted, the risk difference is 20%, so the NNT is 5.) Apply this result with caution! If the NNT is 5, this means that for every five children treated with steroid, one extra admission is prevented. Of the children you see, 75% of them would have gone home anyway.
“Numbers needed to treat” also relate to a specific dichotomous outcome. It's common sense to check that the outcome measures are the same if one is comparing NNTs, but two further issues must be considered, Firstly, the effect of time. One croup trial may have used admission in the following week, and another admission within four hours. Secondly, baseline risks vary between studies. If the croup study was done in the community, perhaps only 10% of the untreated croup patients would have been admitted. Assuming the steroid was equally effective, and reduced the risk of hospitalisation by four fifths, the treated admission rate would be 2%. This gives an NNT of 13.
So the NNT does not reflect the observed benefit of a treatment in practice, but how it compares with another treatment (often placebo). The NNT tells you about one particular population, with a particular degree of baseline risk over a particular period of time. Comparing NNTs requires some form of adjustment for this.
What the NNT does give is a common shorthand. For example, if the NNT to give good outcome, and the NNT to produce side effects can be calculated, the explicit weighing of harm and benefit becomes much clearer.
Structured clinical question
In ventilated preterm babies with pulmonary hypertension and hypoxia [patient], does addition of inhaled nitric oxide [intervention] reduce mortality [outcome]?
Search strategy and outcome
Using Cochrane Library: 1 relevant. Using PubMed—“nitric oxide” AND “premature” AND “randomised clinical trial”: 5 randomised trials. See table 1.
None of the above studies is able to answer our original question. The largest studies were under powered and failed to show important benefits for inNO in preterm babies. The smaller studies showed short term physiological changes; e.g. Subhedar et al: transient fall in OI with 5 ppm inNO; Skimming et al: equal increases in oxygen tension after 15 min of either 5 or 20 ppm inNO.
The possible longer term side effects of inNO are not known. In spite of this, inNO is being routinely used in many UK neonatal units, without the safeguards implicit in participation in a clinical trial (Safeguarding informed parental involvement in clinical research involving newborn babies and infants. A Position Statement. Royal College of Paediatrics and Child Health, December 1999).
(See also Cochrane Library Reviewers' conclusions—which do not contain the two largest studies above, at time of writing, as they are too recent).
CLINICAL BOTTOM LINE
Preterm infants should not be treated with inhaled nitric oxide outside prospective, randomised controlled trials.
Long term follow up of these babies is needed.
Citation Study group Study type (level of evidence) Outcome Key result Comments
Kinsella et al (1999) 80 preterm babies, <34 wk, randomised to 5 ppm inNO or control Double blind RCT, single centre (level 1b) Survival to discharge RR 1.11 95% CI 0.70–1.8 Exp group had imroved PaO2 at 60 min; study underpowered The Franco-Belgium Collaborative NO trial Group (1999) 204 preterm babies, randomised to 10 ppm inNO or control Open, multicentre, RCT (level 1b) Ol at 2 h Median 8.4 (inNO) v 12.4 (control); p=0.005 Greatest in near term infants Study underpowered; baseline Ol higher in the experimental group Subhedar et al (1997) 42 preterm babies (<32 wk) randomised to inNO alone; dexamethasone alone; both; neither Open RCT (level 1b) Death before discharge and/or CLD RR 1.05 95% CI 0.84–1.25 for inNO v controls Neither treatment prevented CLD or death Skimming (1997) 23 preterms randomised to either 5 or 20ppm inNO Open RCT (level 1b) Arterial blood oxygen tension, 15 min after intervention Equal increases in primary outcome, both groups Short term physiological study; no control group