The early trials of inhaled nitric oxide (iNO) in preterm infants were focused on those babies who continued to have major respiratory problems despite antenatal steroids and surfactant, ie, the smallest and sickest infants. The most recent Cochrane review1 includes seven controlled trials2^–8 and showed no significant effect of iNO on mortality, bronchopulmonary dysplasia or risk of intraventricular haemorrhage, although short-term improvements in oxygenation were identified. Follow-up data on babies included in those studies are limited.6 9 10 A follow-up of infants in the single-centre study3 at the age of 2 has reported that 24% of babies treated with iNO had abnormal neurodevelopmental outcomes compared to 46% of controls.9 This significant benefit in the iNO group was primarily due to a 47% decrease in the risk of cognitive impairment. Follow-up of the infants in the study by Subhedar et al2 at the age of 30 months found no significant differences in neurodevelopmental delay (4/7 vs 9/14, RR 0.89, 95% CI 0.37 to 1.75), severe neurodisability (0/7 vs 5/14, p = 0.12) or cerebral palsy (0/7 vs 2/14, p = 0.53) between iNO-treated and control infants.10 A more recent follow-up study, from the National Institutes of Child Health and Human Development (NICHD) study7 indicated no evidence of a difference in the health status of the two groups at a corrected age of 2 years.11

More recent studies using NO for more prolonged periods, utilising the potential of NO beyond simple vasodilatation, are showing more promising results.12 13

Given these somewhat variable results as well as some of the potentially toxic effects of iNO14 and the possibility that these may be more pronounced in preterm infants, it is essential that more longer-term studies be carried out to assess the neurodevelopmental consequences of iNO use.

We have previously reported the results of the INNOVO trial for preterm babies.6 In this study of 108 infants, 59% died and 84% of the survivors had some signs of impairment or disability at the age of 1 year. Twenty per cent of them were classified as severely disabled. A trend towards benefit for iNO on mortality appeared to be outweighed by the suggestion of an increase in impairment or disability in the survivors. However developmental assessment at 1 year often lacks precision regarding the prediction of later health status and hence we were keen to review these children again at a later age.

AIM

The aim of this 4–5-year follow-up of children in the INNOVO trial was to assess the longer-term clinical effectiveness and costs of a policy of adding or not adding iNO to the ventilator gases of preterm neonates with severe respiratory failure.

PATIENTS AND METHODS

The methods for the trial up to the 1-year follow-up have already been reported, and will be briefly summarised here.

Infants of <34 weeks’ gestation, aged less than 28 days and with severe respiratory failure requiring ventilatory support were eligible for trial entry if the responsible clinician was uncertain about whether an infant might benefit from iNO. Infants were randomised either to have iNO added to ventilatory gases, or to ventilatory support without iNO. The suggested starting dose was 5 ppm, doubling to 10 ppm if no satisfactory response was achieved, up to a maximum of 40 ppm. A satisfactory response was defined as an increase in post-ductal arterial oxygen tension (PaO₂) of more than 3 kPa (22.5 mmHg) after the first 15 minutes of giving iNO. A nested randomised study of doses of 5, 10, 20 and 40 ppm found no evidence of a dose–response relationship.15 All other care was left to the discretion of the responsible clinician. Neonatologists were not “blinded” to the allocation.

The study team kept in touch annually with the parents of surviving children by sending birthday cards to the children. Prior to the planned 4–5-year assessment, the child’s general practitioner and health visitor were contacted by the study coordinator to ensure that continued involvement in the trial follow-up programme was considered appropriate. With this agreement, a letter was sent to the family requesting permission to undertake a follow-up visit. For children initially lost to follow-up, an application was made to the Office of National Statistics to ascertain the health authority where children were registered with a general practitioner.

Assessment

The assessment as used in previous trials16 consisted of a medical history obtained from a structured interview with the parents, an assessment of the use of health and educational services and socio-demographic variables using an additional parental questionnaire, a clinical examination which included a neurological assessment, and a functional classification of hearing and vision and a growth assessment. Cognitive development was assessed with the use of the British Ability Scales (BAS) — early years upper level, which measures ability in six core scales.17 These can be used to generate cluster scores (verbal, spatial and pictorial reasoning abilities) and combined to generate a general conceptual ability score, all of which have a standardised mean of 100 and a standard deviation of 15 points. Behaviour was assessed using the strengths and difficulties (S&D) questionnaire completed by the parents.18 This also assessed the impact of behavioural disturbance on the family. We also recorded whether referral had been made to specialist services. The visit lasted 90–120 minutes.

Assessment team

The assessments were performed by two paediatricians trained to perform the various tests, and each covered a different geographical area. Neither was aware of the neonatal course of the child, randomisation, disease course or earlier assessment findings. The families were asked not to inform them about the child’s medical history until the assessment was completed.

Classification of disability

Outcome was assessed in seven domains (cognitive, neuromotor, respiratory, general health, behaviour, vision and hearing and communication). It was classified according to a previously published scheme and was described as normal, impaired or disabled by the degree of functional loss (table 1).19 Children with a test score outside the normal range were defined as impaired, those with some functional loss but who needed little additional support were regarded as having a mild disability, children who used aids and required assistance were classified as moderately disabled, and children who required constant supervision or where special schooling was likely, were classified as severely disabled. A child’s overall status was allocated to the highest degree of disability in any of the seven clinical domains.

Costs at 4-year follow-up

A rigorous cost analysis20 21 was undertaken as part of the 4-year assessment. Information on hospital admissions during the previous year was collected from the hospitals, and on health and community service usage by parent questionnaire. Unit costs were taken from the NHS reference costs database22 and from Netten and Dennet (2002–2003 prices).23

Statistical analyses

These were based on the treatment groups as randomly allocated (“intention to treat”). Those assessed in the follow-up are compared. Comparisons of primary outcomes between treatment groups were presented as relative risks (RRs) with 95% CIs and chi-squared statistical tests for binary variables, and t tests and median tests for continuous data as appropriate. The primary outcome measures were stratified for the major prognostic variables at trial entry: principal diagnosis leading to respiratory distress, post-natal age, and the severity of respiratory disease. Homogeneity of RRs between strata was tested (Mantel–Haenzel chi-square). Confidence intervals of the differences between the groups were reported using non-parametric bootstrap techniques.24 Data were analysed using SAS v8.2 and Stata 9.0.25 26

RESULTS

Recruitment to the trial began in February 1997 and ended as planned in December 2001. A total of 108 infants were recruited (55 allocated to the iNO arm and 53 controls without iNO) from 15 neonatal units in the UK and Republic of Ireland. Forty-four children survived to their first birthday (25 iNO and 19 no iNO). A further child (in the iNO arm) died aged 950 days. The cause of death was given as pneumonia, severe chronic lung disease and extreme prematurity. This child had been classified as severely disabled at 1 year of age. Five further children were not assessed in the 4–5-year follow-up. They were all known to be alive, but in one case in each arm the parents refused; in one case in each arm the parents initially agreed but then did not keep numerous appointments and were eventually considered as de facto refusers; and the fifth child (no iNO) was in foster care and access was not granted. At the 1-year assessment four of these children had been classified as having a “non-severe disability”, and the fifth (the de facto refusal in the no iNO arm) had been classified as “normal” at 1 year.

Follow-up assessments were carried out between September 2003 and January 2005 for 38 children at a median age of 4-and-a-half years (4.63, interquartile range (IQR) 0.84 in the iNO group; 4.52, IQR 0.9 in the no iNO group). Each child was examined at home. Table 2 shows the information available at trial entry, whether or not they received iNO and their overall status at the 1-year follow-up for these 38 children, as well as for all 108 infants recruited. As might be expected, those in the 4–5-year follow-up were more likely to have been of longer gestation, higher birth weight and be less severely ill.

Overall assessment

The proportion of children in each group categorised as having a severe disability, other disability or no disability at 4–5 years were similar to those in the trial at 1 year, and did not differ between the two groups. In all, only 8 children of the original 108 recruited to the trial were classified as normal across all domains at the age of 4–5 years (5 iNO vs 3 no iNO) with a further 5 (3 iNO vs 2 no iNO) having impairment only (table 3). Overall 10 children were classified as having mild disability (6 iNO vs 4 no iNO) and 9 as having moderate disability (5 vs 4). Six children were severely disabled (3 in each group).

Thirty-four of the 55 children in the iNO group (62%) had died or were severely disabled at this follow-up compared to 37/53 (70%) in the no iNO arm (RR 0.89, 95% CI 0.67 to 1.16). Stratifying for post-natal age, principal diagnosis and disease severity produces very similar RRs (all tests of homogeneity p>0.59).

Cognitive assessment

Four (three in the iNO group) children were unable to participate fully in the assessment with the BAS because of severe developmental delay. For those children who were able to participate (n = 19 for iNO and n = 15 for no iNO) the mean general conceptual ability score (GCAS) was similar in the two groups (iNO 91.2 (SD 21.1) vs no iNO 81.3 (SD 22.5)). There was no evidence of any differences between the randomised groups for mean cluster T scores for verbal ability (iNO 95.7 (SD 23.8) vs no iNO 88.3 (SD 21.9)), pictorial reasoning (iNO 99.2 (SD 21.1) vs no iNO 87.7 (SD 21.6)), spatial abilities (iNO 80.7 (SD 23.1) vs no iNO 70.7 (SD 24.8)) or non-verbal composite (iNO 131.5 (SD 31.66) vs no iNO 113.8 (SD 31.7)) — all p>0.1. The tasks for the spatial ability (pattern construction and copying) were performed with the most difficulty, with the mean scores being below the normal range. There was no evidence of any differences in the mean BAS core T scales of verbal comprehension, picture similarities, naming vocabulary, pattern construction, early number concepts and copying between the two groups (all p>0.1).

Neuromotor, sensory and communication assessment

Table 3 shows the degree of disability of the children in the neuromotor, visual and hearing or communication domains. There were three children who had severe disability in neuromotor function, who were all unable to walk without assistance. Two were unable to sit, one had no head control and one was fully dependent for feeding and dressing. No children were completely blind, but one in each group was categorised as seeing only light or gross movement. Other children had less severe impairments: squint was present in 15 children (39%), 10 (45%) in the iNO group and 5 (31%) in the no iNO group; and 11 (50%) in the iNO group and 4 (29%) in the no iNO group wore glasses. One child had hearing impairment not corrected by aids and one corrected with aids (both in the iNO group). In 5 children (13%) there was no recognisable speech (3 iNO vs no iNO).

Behaviour

Although children in the no iNO group appeared more likely to have scores in the “abnormal behaviour” range from the parent S&D questionnaire (eg, overall, 6/22 iNO vs 5/15 no iNO; hyperactivity, 5/22 iNO vs 6/15 no iNO; and conduct, 7/22 iNO vs 8/15 no iNO), none showed evidence of a difference between the randomised groups (all p>0.20). No children reached the criteria for a severe disability on the basis of behaviour. However, seven children were moderately disabled by behaviour problems and had required specialist help (3 iNO vs 4 no iNO) and two (no iNO) had received medication. Parents reported five children in the iNO group and four in the no iNO group having an impact score of 2 or more indicating behavioural difficulties interfering with activities of their daily lives.

Respiratory, general health and health service use

In total 13 children had been diagnosed as having asthma at some point: 9 in the iNO group and 4 in the no iNO group. Of these, 6 of the iNO group and 3 of the no iNO group reported wheezing over the last year, and in addition 6 children (3 in each arm) had reported wheezing in the last year, who had not had a previous diagnosis of asthma. Eleven children had used bronchodilators in the last 12 months (7 in the iNO arm and 4 in the no iNO arm) and 7 children had used inhaled steroids (4 iNO vs 3 no iNO). Eleven children had seen a specialist about breathing problems (7 in the iNO group and 4 in the no iNO group). Eight children had received oxygen therapy at home after discharge from hospital although this had been discontinued at the time of visiting (mean age of stopping 10.8 months (SD 6.9) for the 4 children in the iNO group and 12.8 months (SD 7.9) for the 4 children in the no iNO group). One child had received a past tracheostomy (no iNO group).

Five children had suffered from seizures in the previous 12 months (3 iNO vs 2 no iNO). All but one of these (no iNO) were on regular medication. One child (iNO group) had more than one fit a month and was classed as having a mild disability.

Five children had scarring that was likely to be of cosmetic or functional significance (2 in the iNO group and 3 in the no iNO group). Four children had problems with gastro-oesophageal reflux (1 in the iNO group and 3 in the no iNO group) and two of these had necessitated surgical treatment. Two children (1 in each arm) continued to need a stoma. One child had shunted hydrocephalus (no iNO arm).

There was no evidence of differences between the groups in terms of their weight, height or head circumference, and it was noted that the standardised mean values for weight (−0.86 iNO and −1.02 no iNO), height (−0.9 vs −0.68) and head circumference (−1.48 vs −1.53) were lower than for the normal population.

Overall 29 children (18 iNO vs 11 no iNO) had seen the family doctors in the last 12 months, five (all iNO) had consulted over 12 times. Eleven children (7 iNO vs 4 no iNO) had seen a paediatrician or chest specialist in the last 12 months. Thirty children (18 iNO vs 12 no iNO) had been seen by other health professionals in the last 12 months. Seventeen children had been admitted to hospital for various reasons in the previous year (9 iNO vs 8 no iNO) and for 9 children (3 iNO vs 6 no iNO) this had been more than once. The reasons for each re-admission included respiratory problems (27 iNO (this includes one child who had 25 re-admissions) vs 1 no iNO), surgery (3 iNO vs 9 no iNO) and others (5 iNO vs 7 no iNO).

A number of children were receiving support for special needs, 16 were receiving physiotherapy (9 iNO vs 7 no iNO), 17 were receiving speech and language therapy (9 iNO vs 8 no iNO) and 11 were receiving occupational therapy (6 iNO vs 5 no iNO). Fifteen children were under orthoptists (8 iNO vs 7 no iNO) and 5 children had pre-school teacher counsellors (2 iNO vs 3 no iNO). Nine children were seeing psychologists (4 iNO vs 5 no iNO).

While both groups have sizeable mean costs at 4 years, the difference between the groups in resource use and cost is small and not statistically significant (table 4).

DISCUSSION

This follow-up study did not demonstrate that the short-term use of iNO for severe respiratory failure in preterm infants provides either benefit or harm in terms of longer-term clinical effectiveness. This is the only trial of preterm babies with follow-up to 4–5 years and uses functional outcomes that matter to families. The findings are unlikely to be due to selection bias, as there was secure random allocation generating two comparable groups in the two trial arms. There was 12% loss to follow-up (5 of the 43 surviving children). In addition there was a standardised assessment by two people who were not connected with the initial care and without knowledge of treatment allocation or the previous health status of the child. However, the numbers are small and the treatments were not blind during the babies’ stay on the neonatal unit, which means that there could be a dilution of effect (crossovers, concomitant treatments).

The findings of this study do not show there was a difference in the longer-term outcome between those babies treated with iNO and those that were not. The findings are compatible with those of the NICHD trial where there was no difference in neurodevelopmental impairment among survivors at 18–22 months,12 and impairment was present in 47% of iNO survivors and 46% of controls. The trial in which iNO treatment was associated with improved neuro-developmental outcomes at 2 years of age9 included a population of preterm babies who had less severe lung disease than those babies in the INNOVO trial and the NICHD trial. It is possible that the babies that were treated in the INNOVO trial were so sick to start with that the potential for benefits was limited. Future studies therefore need to concentrate on determining what premature population might be appropriate for treatment with iNO.

Further studies published recently reporting short-term outcomes have aimed to address this issue12 13 and an ongoing trial in European centres will also provide further data. Both of these new published trials focused on babies with markedly less severe lung disease than those included in either INNOVO or the NICHD trial. Both studies showed some clinically important improvements in the treated babies but in both cases these represented improvements in a subset rather than an across-the-board benefit for the treated group of babies.

A full economic assessment at 1 and 4 years was incorporated into the initial design of the study. We were unable to demonstrate that health gains at 4 years offset the extra costs associated with iNO at 1 year.6 Although the numbers are small this cost analysis adds to the evidence suggesting that iNO used in this way and in this group of babies is unlikely to be a cost-effective intervention for severe respiratory failure in preterm infants.

Given our findings the challenge remains to identify those premature babies whose response to NO leads to health improvements that justify these additional costs. At the moment that information remains elusive.

The INNOVO study also looked at a population of term infants and the long-term outcome of these babies is currently being analysed.