NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Allen MC, Donohue P, Gilmore M, et al. Inhaled Nitric Oxide in Preterm Infants. Rockville (MD): Agency for Healthcare Research and Quality (US); 2010 Oct. (Evidence Reports/Technology Assessments, No. 195.)

Cover of Inhaled Nitric Oxide in Preterm Infants

Inhaled Nitric Oxide in Preterm Infants.

Show details

3Results

Literature Search/Abstract/Article Review

The literature search process identified 3104 unique citations. During the abstract review process, we excluded 2650 citations that did not meet one or more of the eligibility criteria (see Chapter 2 for details). At article review, we excluded an additional 423 articles that did not meet one or more of the eligibility criteria (see Appendix D, Excluded Articles). Thirty-one articles were eligible for inclusion in the review. There were 14 RCTs described in 23 articles, and eight observational studies (Figure 2).

This figure describes the sources used to identify potentially relevant articles for this review. It also describes the review process by showing how many articles were identified as potentially eligible, and reasons for ineligibility.

Figure 2

Summary of literature search (number of articles). * Total may exceed number in corresponding box, as articles excluded by two reviewers at this level.

Description of the Types of Studies Retrieved

There were 14 articles that applied to Key Question 1. Fourteen articles (all RCTs) applied to Key Question 2. Twelve articles applied to Key Question 3; three original RCTs with six followup studies, and three observational studies. Seventeen articles applied to Key Question 4; six RCTs with four followup studies, and seven cohort studies. Twenty-one articles applied to Key Question 5; 14 RCTs with seven followup studies (Table 1). Eight studies containing 13 case reports were also examined. As mentioned in Chapter 2, we reviewed the case reports and determined that the data were not of sufficient detail to be considered further.

Table 1. Included articles.

Table 1

Included articles.

Risk of Bias

As described in Chapter 2, we assessed each individual study for risk of bias using a study design specific tool. (See Appendix E, Evidence Tables 1 and 2.)

The RCTs and their followup studies were not assessed separately for risk of bias. Six of the 14 RCTs (along with their five followup studies) were assessed as having low risk of bias.30, 34, 36, 37, 39, 40, 44, 56, 57, 58, 59 Three of the RCTs were determined to be at fair risk of bias.60-62 The remaining five RCTs were assessed as having a high risk of bias.63-67 Figure 3 provides a summary of the risk of bias for the RCTs. (Appendix E, Evidence Table 1).

This figure identifies the number of RCTs that are given a positive (low risk of bias), negative (high risk of bias), or unclear risk of bias score for the following categories: adequate sequence generation, allocation concealment, blinding of project personnel, blinding outcome assessors, adequately reported incomplete data, no selective outcome reporting, free of other risks of bias.

Figure 3

Summary of risk of bias for RCTs.

None of the eight observational studies were considered to have a low risk of bias. Five received a fair risk of bias assessment, and received this score for a variety of reasons.38, 68-71 Three observational studies were assessed at high risk of bias.72-74 (Appendix E, Evidence Table 2).

Key Question 1. Does inhaled nitric oxide (iNO) therapy increase survival and/or reduce the occurrence or severity of bronchopulmonary dysplasia (BPD) among premature infants who receive respiratory support?

Major Findings

  • There is no statistically significant effect of iNO compared to placebo on survival or mortality rates in preterm NICU infants requiring mechanical ventilation.
  • There is insufficient evidence to determine whether iNO reduces the rate of bronchopulmonary dysplasia (BPD), as defined by requiring supplemental oxygen at 36 weeks postmenstrual age, in preterm NICU infants requiring mechanical ventilation.
  • There is a small but statistically significant reduction in the composite variable of death or BPD at 36 weeks PMA for infants treated with iNO compared to infants in control groups.
  • Preterm infants who required mechanical ventilation and were the subjects of randomized controlled trials of inhaled nitric oxide were a high risk population with high mortality and BPD rates during NICU hospitalization.

Detailed Analysis

We identified 14 RCTs that compared treatment with iNO to standard treatment in preterm infants requiring mechanical ventilation (Table 2). They varied as to inclusion and exclusion criteria; age of enrollment; dose, timing and duration of iNO; and outcome variables reported. Current labeling of iNO is for use in infants born after 34 weeks gestation with respiratory failure, so we included two RCTs of preterm infants born at or before 34 weeks gestation.59, 79 All but one RCT began enrollment and started iNO during the first week after birth; the RCT that differed from the others enrolled infants and started iNO at seven to 21 days after birth.34 The 14 RCTs varied widely as to severity of respiratory illness, birth weight (BW), gestational age, chronological age from birth, and postmenstrual age (PMA, gestational age plus chronological age, a proxy for degree of prematurity) when treatment was initiated (Table 3). Their study designs varied widely in terms of dose (5 to 40 parts per million (ppm)), duration (1 to 24 days), and mode of delivery. The 14 RCTs varied so widely that it was difficult to group them together in a way that took these important variables into account. For Key Question 1 (and Key Question 2), we viewed the aggregation of these 14 RCTs as providing data on a continuum of exposures to iNO (as listed above). This discussion of death and BPD includes all 14 RCTs that provide data for the variables we were charged with systematically reviewing: death or survival, BPD and the composite variable of death or BPD at 36 weeks PMA. Key Questions 4 and 5 explore data regarding subgroups and variations in administration of iNO, respectively.

Table 2. Summary of outcomes for RCTs addressing KQ1.

Table 2

Summary of outcomes for RCTs addressing KQ1.

Table 3. Study design of randomized controlled trials of inhaled nitric oxide in preterm infants.

Table 3

Study design of randomized controlled trials of inhaled nitric oxide in preterm infants.

Each of the 14 RCTs reported mortality data (three reported only death at 7 or 28 days), and 11 reported data on BPD (Table 2). Three studies focused on changes of oxygenation index (OI) at 2 to 24 hours after starting iNO therapy.60, 61, 65 Six RCTs reported using placebo gas in the control group and keeping NICU staff masked as to study arm assignment.34, 37, 39, 40, 58, 62 There were four multicenter RCTs and one single center RCT that had at least 100 infants per study arm.37, 62 The findings of the 14 RCTs are discussed below by outcome variable: death/survival, BPD rate, severity of BPD and the composite variable of death or BPD (Appendix E, Evidence Tables 3 and 4; Table 2 and 3).

Survival and death. Each of the 14 RCTs reported mortality data, but there was some variation as to the time point cutoff they used for reporting death (e.g., 7 days, 28 days, 36 weeks PMA, death while in the NICU, or death while in the NICU or in the first 365 days for infants who had prolonged NICU stays). We assumed that death occurred in the NICU in the few studies that did not specify the time point cutoff they used for reporting death as they reported only NICU outcomes. One pilot study for a larger single center RCT focused on toxicity of iNO, and reported only death in the first seven days.66 One study that focused on physiologic response to iNO reported death at seven and 28 days.61 For the INNOVO RCT, Field, 2005 reported death in the first day after birth, at two to six days, at seven to 27 days and at 28 days to one year corrected for degree of prematurity.63 We included these three RCTs in the Evidence Table for death (Appendix E, Evidence Table 5). No matter how the 14 RCTs defined and reported death or survival, none of the 14 RCTs reported a statistically significant difference between iNO and control groups.

Our discussion and meta-analysis focuses on the 11 RCTs that report death by 36 weeks PMA or in the NICU, a specified variable we were asked to include in our systematic review. Six of the RCTs used the composite variable, death or BPD, as their primary outcome variable,37, 39, 40, 58, 64, 67, and two RCTs used survival without BPD at 36 weeks PMA as their primary outcome variable.34, 62 All three RCTs, Su 2008, Hascoet, 2005, and Srisuparp, 2002, that focused on immediate physiologic found no significant differences in mortality between the iNO and control groups.60, 61, 65 Survival to NICU discharge was the primary outcome in only one study, an early multicenter RCT published by Kinsella,1999.59 A lower than expected recruitment rate, in combination with an interim analysis that suggested they were unlikely to find a difference in survival, prompted them to terminate this trial early. The survival rates for the iNO group and controls were 25 percent versus 20 percent, respectively with a Relative Risk (RR) of 1.11 (95 percent confidence interval 0.70, 1.80). To make their data comparable to the other RCTs, we report not survival to NICU discharge, but their NICU mortality rates (i.e., 75 percent versus 80 percent) (Appendix E, Evidence Table 5).

The largest multicenter RCT (the EUNO trial), Mercier, 2010, was conducted at 36 centers in nine European countries.62 They enrolled 800 preterm infants born between 24 weeks and 28 6/7 weeks gestation with a BW at or above 500 g who were treated with surfactant and mechanical ventilation or continuous positive airway pressure (CPAP) for respiratory distress syndrome. Within 24 hours after birth, infants were enrolled and treated with low dose iNO (5 ppm) or placebo gas for a minimum of seven days. There was no statistically significant difference in survival at 36 weeks PMA between the iNO group and controls, 86 percent versus 90 percent, respectively, RR 0.74 (0.48, 1.15), adjusted for gestational age, baseline severity of illness, mode of ventilation and country. For comparison with the other RCTs, we present their data in Table 3 and our meta-analyses in terms of death by 36 weeks PMA, 14 percent versus 10 percent for iNO and control groups, respectively.

The second largest multicenter RCT, published by Kinsella, 2006, enrolled 793 preterm infants born at or before 34 weeks gestation with birth weight below 1250 g on mechanical ventilation in the first two days after birth.37 Study infants were treated with low dose iNO (5 ppm) or placebo gas for 21 days or until extubation. The NICU mortality rates were 20 percent in the iNO group and 25 percent in the placebo group, RR 0.79 (0.61, 10.3) (p-value = 0.08) (Appendix E, Evidence Table 5).

Ballard, 2006, enrolled 582 infants born before 33 weeks gestation or less with birth weight below 1250 g who were on positive pressure ventilation (ventilator or CPAP) beyond the first week, seven to 21 days after birth.34 Since preterm mortality is highest during the first week after birth, this RCT had the lowest mortality rates at 36 weeks PMA of all the RCTs of iNO in preterm infants, 5.4 percent in the iNO group and 6.3 percent in placebo controls. Mortality rates were only slightly higher after term, at 44 weeks PMA (6.9 percent versus 6.8 percent) (Appendix E, Evidence Table 5).

In the NICHD Neonatal Research Network’s two RCTs of preterm infants born before 34 weeks gestation (BW 401 to 1500 g in the 2005 study and BW > 1500 g in the 2007 study), Van Meurs reported death before discharge to home or 365 days for preterm infants still hospitalized.39, 40 For both RCTs, the Data Safety and Monitoring Committee noted higher than expected mortality rates, and recommended lowering the OI inclusion criteria. Despite this change to include infants with less severe respiratory failure, mortality rates for preterm infants with BW 400 to 1500 g were 52 percent for the iNO group versus 44 percent for placebo controls (RR 1.16 (0.96, 1.39)), adjusted for study center, BW group, and OI group.40 As might be expected, mortality rates were somewhat lower for infants with birth weight above 1500 g, 36 percent in the iNO group versus 27 percent in controls (RR 1.26 (0.47, 3.41)) when adjusted for OI stratum.39 (Appendix E, Evidence Table 5).

In the largest single center RCT with 207 infants born before 34 weeks gestation with BW below 2000 g, Schreiber, 2003 found no differences in death in the NICU or by six months: 15 percent for iNO versus 23 percent for placebo gas controls (RR 0.68 (0.38, 1.2)), adjusted for type of ventilation.58 There were no statistically significant differences in NICU mortality in the Franco-Belgium, 1999 RCT (27 percent in the iNO group and 35 percent in controls), despite having an initial OI that was higher in the iNO group (median 20.2, interquartile range (IQR) 8.3) than in the control group (median 18.0, IQR 7.4) 60 Neither of two small single center RCTs that each enrolled 40 to 42 infants found statistically significant differences between the iNO and control groups: mortality was 20 percent versus 30 percent, respectively, as reported by Dani, 2006, and 50 percent versus 32 percent with RR 1.57 (0.76, 3.38) as reported by Subhedar, 199764, 67 (Appendix E, Evidence Table 5).

There were no statistically significant differences (neither increase nor decrease) in NICU mortality rates with iNO in any of the individual fourteen RCTs that reported on death or survival. A meta-analysis of the 11 RCTS that reported death by 36 weeks PMA or in the NICU found no statistically significant differences between the iNO and control groups (RR 0.97 (0.82, 1.15)) (Figure 4).

This figure is a forest plot of 11 RCTs used to calculate whether pooled data on the impact of iNO on preterm infants’ death at 36 weeks PMA was significant.

Figure 4

Meta-analysis of 11 studies describing death at 36 weeks PMA or in the NICU.

The relatively high mortality rates for infants enrolled in most of these RCTs are striking. Regardless of group assignment, seven studies reported mortality rates as high as 25 percent to 65 percent,39, 40, 59-61, 63, 64 and another four reported mortality rates of 15 to 30 percent.37, 58, 65, 67 The Ballard, 2006 RCT reported the lowest mortality rates (5.4 percent for the iNO group and 6.3 percent for controls).34 Most preterm infants who die in the NICU die during the first week after birth. The low mortality rates reported in Ballard, 2006 reflects a difference of study design in that they enrolled preterm infants on mechanical ventilation or CPAP at one and to three weeks after birth. We therefore removed the data from the Ballard RCT, and repeated the meta-analysis for the remaining ten RCTs. The results did not change: RR 0.98 (0.81, 1.17) (Figure 5). This sensitivity analysis confirmed no statistically significant effect of iNO compared to control on NICU death or survival to discharge from the NICU in preterm infants requiring positive pressure ventilation.

This figure is a forest plot of 10 RCTs used to calculate whether pooled data on the impact of iNO on preterm infants’ death at 36 weeks PMA was significant. One RCT was removed for a sensitivity analysis; this RCT was sufficiently different in design to merit this additional analysis.

Figure 5

Meta-analysis of 10 studies describing death at 36 weeks PMA or in the NICU without Ballard, 2006.

Bronchopulmonary dysplasia. The term bronchopulmonary dysplasia was introduced in 1967 when Northway reported a case series of preterm infants with respiratory distress syndrome (RDS) who developed a chronic lung disease with characteristic radiographic and pathologic features.82 Although it has always been defined clinically, the definition of BPD has evolved with neonatal intensive care.83 The clinical, radiographic and pathologic features of BPD have changed as new technologies, medications and management strategies have been introduced, leading to dramatic reductions in gestational age specific neonatal mortality and a lowering of the limit of viability to now 22 to 24 weeks gestation. The evolution of BPD is reflected in its various definitions, including definitions based on persistent respiratory symptoms, radiographic features, and treatments (e.g., requiring supplemental oxygen at 28 days from birth, or a more severe BPD, requiring oxygen at 36 weeks PMA).

Twelve RCTs provide data on BPD at 36 weeks PMA, but there was some variation in how each RCT defined BPD. Six RCTs defined BPD simply as requiring supplemental oxygen at 36 weeks PMA.39, 40, 59, 60, 63, 67 One multicenter study37 and three single center studies58,64,65 refined the definition of BPD by adding the requirement of radiological evidence of BPD. Although there is general agreement that infants on mechanical ventilation or supplemental oxygen above 30 percent FiO2 at 36 weeks PMA have BPD, some question whether infants on low flow nasal cannulas with a FiO2 of 30 percent or less should be included in the BPD at 36 weeks PMA category.

Walsh published in 2003 an algorithm for physiologic BPD at 36 weeks PMA that includes an oxygen challenge test for infants on less than 30 percent FiO2.84 Four RCTs used these criteria for categorizing BPD at 36 weeks PMA.34, 39, 40, 62 Van Meurs, 2005 and 2007 reported that the rate of BPD as defined by Walsh was somewhat lower than the rate of BPD defined as on supplemental oxygen at 36 weeks PMA.39, 40 In the NICHD trial of infants born before 34 weeks gestation with birth weight below 1500 g, physiologic BPD rates, 50 percent in the iNO group and 60 percent in controls, RR 0.87 (0.68, 1.10), were lower than rates of BPD defined as requiring oxygen at 36 weeks PMA, 60 percent versus 68 percent, RR 0.90 (0.75, 1.08) (both RR were adjusted for study center, BW group and OI group).40 In the NICHD RCT of infants with birth weight above 1500 g, the physiological definition classified one more infant treated with iNO as having BPD, and one less control infant as having BPD, resulting in BPD rates of 36 percent versus 40 percent respectively, RR 0.74 (0.26, 2.09) adjusted for OI stratum.39 (Appendix E, Evidence Table 6). Using the physiologic definition at 36 weeks PMA, Mercier, 201062 reported lower BPD rates, 24 percent in the iNO group compared to 27 percent in controls, RR 0.84 (0.58, 1.17), adjusted for gestational age, baseline severity of illness mode of ventilation, and country. Their inclusion criteria differed from two RCTs reported by Van Meurs39, 40 (Table 3) in their focus on gestational age (i.e., gestational age 24 to 28 6/7 weeks) rather than BW and lower severity of initial illness (mechanical ventilation with FiO2 at or above 30 percent).

The twelve RCTs also vary as to the denominator used when calculating rate of BPD at 36 weeks PMA: Five used the total number of infants in each group,34, 63-65, 67 and seven RCTs used the number of survivors in each group.37, 39, 40, 58-60, 62, 85 The small single center Subhedar, 1997 RCT reported BPD rates both for the total group (50 percent for the iNO group versus 64 percent for controls) and for survivors (100 percent versus 90 percent).64

Dani, 2006 noted that infants treated with iNO had half the rate of BPD at 36 weeks PMA than the controls (30 percent versus 60 percent, respectively, p-value = 0.067, BPD rate for the total group).67 An unplanned interim analysis revealed a statistically significant reduction in their primary outcome, death or BPD (p-value = 0.016). On the recommendation of their consulting statisticians and two independent observers, the study was terminated early, with enrollment of only 40 of the anticipated 52 infants. Another small single center RCT found no statistically significant differences in rate of BPD at 36 weeks PMA (31 percent for the total iNO group and 33 percent for the total control group).65 Field, 2005 reported that 26 of 55 infants in the iNO group and 15 of 53 infants in the control group had BPD at 36 weeks PMA, 47 percent versus 28 percent. Ballard, 2006 reported rates of BPD at 36 weeks PMA for the total iNO group as compared with the total placebo gas control group: 50.7 percent versus 56.9 percent, respectively (Appendix E, Evidence Table 6).

In addition to the Mercier and two Van Meurs NICHD Neonatal Research Network RCTs, Kinsella, 2006 reported BPD rates using as denominator the number of infants alive at 36 weeks PMA37, 39, 40 There was no statistically significant differences between the iNO and placebo gas control groups, 65 percent versus 68 percent, respectively, RR 0.96 (0.86, 1.09). Using the number of survivors as denominator, Kinsella, 1999 and Schreiber, 2003 reported differences in rates of BPD at 36 weeks PMA that were not statistically significant.58 Kinsella, 1999 reported that 60 percent of survivors in the iNO group had BPD at 36 weeks PMA as compared with 80 percent of control survivors.59 Schreiber, 2003 reported that 39 percent of iNO group survivors compared to 53 percent of control group survivors had BPD at 36 weeks PMA, RR 0.74 (0.53, 1.03) (Figure 6).58 (Appendix E, Evidence Table 6).

This figure is a forest plot of eight RCTs used to calculate whether pooled data on the impact of iNO on BPD at 36 weeks PMA, in preterm infants, is significant.

Figure 6

Meta-analysis of eight studies describing BPD at 36 weeks PMA among survivors.

Despite variations in how BPD was defined and calculated, there were no statistically significant differences in rates of BPD at 36 weeks PMA between the iNO group and controls in any of the RCTs. Subhedar, 199764 demonstrated how drastically BPD rates can differ when they are calculated using survivors as compared with the total group as denominator. For this reason, we did not do a meta-analysis with all 12 RCTs. We included eight studies in a meta-analysis of the rate of BPD at 36 weeks PMA in survivors. The small difference was not statistically significant (RR 0.93 (0.86, 1.003)) (Figure 6) (Appendix E, Evidence Table 6).37, 40, 58, 59, 64

Other measures of severity of bronchopulmonary dysplasia. Although the most accepted BPD definition is based on being on supplemental oxygen at 36 weeks PMA, there are other measures of severity of lung disease (e.g., duration of mechanical ventilation and oxygen supplementation, treatment with medications for lung disease), and other time points for reporting the need for mechanical ventilation of supplemental oxygen (e.g., at 40 weeks PMA, 44 weeks PMA and NICU discharge). We found no RCTs that reported number of infants on mechanical ventilation at 36 weeks PMA.

Ballard, 2006 reported statistically significantly fewer infants in the iNO group than controls remained in the hospital, and on mechanical ventilation, nasal continuous positive airway pressure (CPAP) or supplemental oxygen at 40 weeks PMA (p-value = 0.01), and at 44 weeks PMA (p-value = 0.03).34 At 40 weeks PMA (i.e., full term), six percent in the iNO group and 10 percent of controls were hospitalized and on mechanical ventilation, and 22 percent versus 29 percent were hospitalized and on supplemental oxygen. Kinsella, 1999 reported that 54 percent of infants in the iNO group were discharged home on oxygen as compared with 80 percent of control infants, RR 0.65 (0.41, 1.02).59 In contrast, only nine percent of all infants in each group were discharge home on supplemental oxygen in Field, 2005.63

Kinsella, 2006 reported no differences between the iNO group and controls in proportion of infants ever treated with postnatal corticosteroids (60 percent versus 56 percent).37 There were no statistically significant differences in proportion of survivors at 36 weeks PMA who were on bronchodilators (20 percent versus 20 percent), corticosteroids (15 percent versus 12 percent) or diuretics (37 percent versus 38 percent). In Franco-Belgium, 1999 there were also no statistically significant differences between the 29 survivors in the iNO group who were treated with steroids (54 percent versus 72 percent) or beta-mimetics (21 percent versus 39 percent) than the 29 control survivors.63 Field, 2006 reported that 40 percent and 34 percent of the iNO versus control group were treated with corticosteroids63 (Appendix E, Evidence Table 6). Eight RCTs reported mean duration of supplemental oxygen, mechanical ventilation or CPAP.39, 40, 58, 60, 62, 63, 65, 67 Dani, 2006 reported a statistically significant lower mean duration of supplemental oxygen reached statistical significance for all infants in the iNO compared to all in the control group (47.3+/-39.4 versus 69.4+/-30.2, p-value = 0.05), but no statistically significant differences in mean days of mechanical ventilation or CPAP.67 Two other RCTs found no statistically significant differences between the total iNO group and controls in mean duration of mechanical ventilation60, 65 nor mean duration of supplemental oxygen.60 (Appendix E, Evidence Table 6). The largest multicenter RCT published in 2010 by Mercier reported no statistically significant differences in mean duration of mechanical ventilation between the iNO group and controls, 44+/-26 versus 45+/-29, respectively, p-value = 0.68, but did not specify whether these data were for the total groups or survivors.62 Three RCTs reported mean duration of supplemental oxygen or mechanical ventilation in survivors.39, 40, 58 Van Meurs, 2007 RCT of preterm infants with birth weight above 1500 g, the mean duration of mechanical ventilation was 8.7+/-5.4 days for the nine survivors in the iNO group and 16.8+/-13.9 for the 11 controls (p-value = 0.08).39 In their RCT of preterm infants with birth weight 400 to 1500 g, there were no statistically significant differences between the iNO and control groups in mean duration of mechanical ventilation (39+/-45 versus 47+/-53) or supplemental oxygen (84+/-63 versus 91+/-61).40 In Schreiber, 2003, the median duration of mechanical ventilation was 16 days for the iNO group (the interquartile range was 8 to 48) and 28.5 days (IQR 8 to 48) for controls p-value = 0.19.58 (Appendix E, Evidence Table 6).

As a part of their analyses of costs and resource utilization, Field, 2005 reported data regarding mechanical ventilation and supplemental oxygen for infants who survived and for the total group.63 Median (interquartile range) for days on mechanical ventilation after randomization was 7.0 (2.0, 28.0) for all infants in the iNO group versus 4.0 (1.0, 9.0) in all controls, and 15.0 (6.0, 28.0) for survivors in the iNO group versus 12.0 (5.0, 36.0) in surviving controls. The data for days on supplemental oxygen after randomization were similar.63 (Appendix E, Evidence Table 6).

Of the eight RCTs that reported various measures of severity of BPD, only two reported differences between the iNO and control groups that approached statistical significance, and both favored iNO. Ballard, 2006 reported a statistically significant reduction in hospitalization and respiratory support at 40 and 44 weeks PMA with iNO (p-value = 0.01 and p-value = 0.03, respectively).34 Dani, 2006 reported a lower duration of supplemental oxygen with iNO (p-value = 0.05).67 There are insufficient data to perform a meta-analysis for any measure of severity of BPD due to lack of uniformity in definitions used. Although a number of RCTs reported duration of mechanical ventilation and/or supplemental oxygen, they varied as to whether they used mean +/- standard deviation or median (interquartile range), and whether the data were calculated for the total group or only for survivors.

Death or bronchopulmonary dysplasia at 36 weeks PMA. The composite outcome of death or BPD at 36 weeks PMA was reported in 11 RCTs: it was the primary outcome variable for six RCTs 39; its complement, survival without BPD at 36 weeks PMA, was the primary outcome variable in the Mercier, 2010 RCT34, 62; in two RCTs the primary variable was OI at a specified time60, 65; in one RCT the primary outcome variable34, 37, 40, 58, 64 was survival to discharge from the NICU59; and for one RCT the primary outcome variable was death or severe neurodevelopmental impairment.59, 63 In one multicenter RCT and two single center RCTs, there were statistically significant differences between the iNO group and controls in the composite outcome of death or BPD.34, 58, 67 All eleven RCTs were included in our meta-analysis. (Appendix E, Evidence Table 7).

Ballard, 2006 found a statistically significant benefit in their primary outcome, survival without BPD at 36 weeks PMA, for the iNO group compared to placebo controls, 44 percent versus 37 percent, RR 1.23 (1.01, 1.51).34 The number needed to treat was 14. Although their study sample was similar to other RCTs (birth before 33 weeks gestation with birth weight at or below 1250 g), infants were enrolled later than in other studies (at 7 to 21 days, compared to within the first week), and the minimum duration of treatment for the Ballard study was 21 days. For comparison with the other RCTs, we used the complement composite variable, rates of death or BPD at 36 weeks PMA, 56 percent of the iNO group versus 63 percent of the placebo control group) in Appendix E, Evidence Table 7 and Figure 7.

This figure is a forest plot of 11 RCTs used to calculate whether pooled data on the impact of iNO on death or BPD at 36 weeks PMA, in preterm infants, is significant.

Figure 7

Meta-analysis of studies describing death or BPD at 36 weeks PMA.

Schreiber, 2003, the largest single center trial, reported a statistically significant difference in rate of death or BPD.58 In the iNO group (n = 105), 49 percent died or developed BPD compared to 64 percent in the placebo control group (n = 102), RR 0.76, (0.60, 0.97). This RCT enrolled infants born before 34 weeks gestation as other RCTs but with birth weight below 2000 g, and they treated study infants with iNO for seven days (Appendix E, Evidence Table 7).

The other single center RCT that found a statistically significant difference between the iNO group and controls in the outcome of death or BPD was reported by Dani, 2006.67 This RCT was stopped early (n = 40) because an unplanned interim analysis found a statistically significant difference (p-value = 0.02) in death or BPD, their primary outcome. Only 50 percent of infants in the iNO group died or developed BPD, compared to 90 percent of infants in the control group, RR 0.11 (0.02, 0.61). In this study, the controls were not treated with placebo gas but received standard care and NICU staff was not masked as to study status. The mean duration of treatment with iNO was 98.5 +/- 21.4 hours (4.1 days) (Appendix E, Evidence Table 7).

The largest multicenter RCT published in 2010 by Mercier reported no statistically significant difference between 395 infants in the iNO group compared to 400 in the placebo gas control group in their primary outcome variable, survival without BPD at 36 weeks PMA.62 They used low dose 5 ppm iNO for seven to 21 days and the physiologic definition of BPD, as published in 2003 by Walsh.57 Sixty-five percent of the infants in the iNO group and 66 percent of infants in the placebo gas control group survived without BPD at 36 weeks PMA, RR 1.05 (0.78, 1.43). For comparison with other RCTs, we use the complement combined variable death or BPD at 36 weeks PMA, 35 percent versus 34 percent, respectively (Appendix E, Evidence Table 7 and Figure 7.

Just as they found no statistically significant differences in mortality or BPD rates, the two Van Meurs Neonatal Research Network RCTs, the large multicenter Kinsella, 2006 RCT, and Subhedar’s small single center RCT found no statistically significant differences in the composite variable of death or BPD at 36 weeks PMA.37, 39, 40, 64 Both NICHD trials were terminated at the second interim data analysis of this study, at the recommendation of their data safety monitoring committee, based on no statistically significant differences in death or BPD and concerns about significantly higher rates of severe intracranial hemorrhage or periventricular leukomalacia (PVL) in the larger RCT.39, 40 Rate of death or BPD at 36 weeks PMA was 80 percent for the iNO group and 82 percent for controls, RR 0.97 (0.86, 1.06) adjusted for study center, birth weight group and OI group.40 The NICHD trial of infants birth weight above 1500 g reported that rate of death or BPD at 36 weeks PMA was 50 percent for the iNO group and 60 percent for controls, RR 0.80 (0.43, 1.48) adjusted for OI.39 The rate of death of death or BPD in the large Kinsella, 2006 multicenter RCT was 72 percent in the iNO group compared to 75 percent in controls, RR 0.95 (0.87, 1.03).37 Kinsella, 1999, a trial that included infants with more severe respiratory failure, reported much higher rates of death or BPD at 36 weeks PMA, 77 percent versus 91 percent, RR 0.85 (0.70, 1.03), but no significant differences between groups.59 Subhedar, 1997 reported even higher rates of death or BPD at 36 weeks PMA, 95 percent in the iNO group and 100 percent in controls, RR 1.04 (0.92, 1.19).64 (Appendix E, Evidence Table 7).

Two RCTs focused on early physiologic response to the administration of iNO gas. They both had oxygen index (OI) as their primary outcome variable, and differed only as to timing. Franco-Belgium, 1999 found no statistically significant differences in OI at two hours after administration of iNO,60 whereas Su, 2008 reported an OI at 24 hrs after administration of iNO that was statistically significantly lower in the iNO group.65 Rates of the composite variable, death or BPD at 36 weeks PMA, in the iNO versus control groups were 45 percent versus 53 percent, respectively, for the Franco-Belgium, 1999 and 50 percent versus 64 percent, respectively, for Su, 2008.

Our meta-analysis of pooled data from all 11 RCTs for death or BPD at 36 weeks PMA found a small but statistically significant difference in favor of iNO, RR 0.927 (0.870, 0.988) (Figure 7). It has been suggested that the study by Ballard, 2006,34 should not be included in meta-analyses as it had a very different study design as well as the lowest mortality rates when compared to the other RCTs. In a sensitivity analysis, removing Ballard, 2006 from this meta-analysis did not change the effect estimate (RR 0.93). However, not surprising given the size of this study, removing it from the analysis did influence the confidence intervals; the confidence interval for the meta-analysis without Ballard, 2006 included 1 (0.87, 1.000). Running the analysis without Ballard, 2006 did not reduce the statistical heterogeneity, as measured by I2 (Figure 8).

This figure is a forest plot of 11 RCTs used to calculate whether pooled data on the impact iNO on death of BPD at 36 weeks PMA, in preterm infants, is significant. One RCT was removed for a sensitivity analysis; this RCT was sufficiently different in design to merit this additional analysis.

Figure 8

Meta-analysis of 10 studies describing death or BPD at 36 weeks PMA, without Ballard, 2006.

Conclusion

Neither our meta-analysis nor any of the fourteen RCTs found any statistically significant differences in death in the NICU or survival to NICU discharge with iNO. Similarly, there were no statistically significant differences in any of the 12 RCTs that reported rates of BPD at 36 weeks PMA. Our meta-analysis of eight RCTs that reported rate of BPD at 36 weeks PMA for survivors did not find a statistically significant difference between the iNO or control groups, though most of these studies favored the iNO group. Two of eight RCTs that reported other pulmonary outcomes reflecting severity of BPD reported statistically significant findings in favor of iNO: a reduction in rates of hospitalization and respiratory support at 40 and 44 weeks PMA,34 and a statistically significant reduction in mean duration of supplementary oxygen.67 Three of 11 RCTs reported a statistically significant reduction of the composite variable, death or BPD at 36 weeks PMA or its complement, improved survival without BPD at 36 weeks PMA.34, 58, 67 There was a small but statistically significant reduction in favor of iNO in our meta-analysis of all 11 RCTs that reported data for the composite variable, death or BPD at 36 weeks PMA. Ballard, 2006 is considered by some as different from the other studies in terms of study design (i.e., not enrolling or initiating treatment until a week or more after birth, and a minimum treatment duration of 21 days), and it had the lowest mortality rate of all 14 RCTs. Excluding data from the Ballard, 2006 and rerunning the meta-analysis resulted in the same effect estimate but a wider confidence interval that included 1. A meta-analysis with all 11 trials may provide a more complete picture of the available evidence, when considering the effect of iNO in a continuum of exposure at various postmenopausal ages. When death or BPD at 36 weeks PMA is viewed in terms of its complement, the pooled estimate of risk favors iNO with a small but statistically significant improvement in survival without BPD at 36 weeks PMA by seven percent. This finding leads to questions about short term risks, longer term neurodevelopmental, pulmonary and other health outcomes, whether iNO is more effective in certain subgroups, and optimal doses, and methods of drug administration, which are discussed in Key Questions 2, 3, 4 and 5.

Key Question 2. Are there short term risks of iNO therapy among premature infants who receive respiratory support?

Major Findings

  • There is insufficient evidence of a neuroprotective effect of iNO in preterm infants.
  • There is no evidence that treatment of preterm infants with iNO influences the rates of other complications of prematurity, including patent ductus arteriosus (PDA), sepsis, necrotizing enterocolitis (NEC), severe retinopathy of prematurity (ROP), pulmonary hemorrhage, or air leaks.
  • No study reported accumulation of toxic levels of methemoglobin or nitrogen dioxide.

Detailed Analysis

Preterm birth requires infants to utilize organ systems that are not yet fully mature.86 The many complications of prematurity are multifactorial in etiology, but the highest risk factor is degree of prematurity. Infants born at 22 to 23 weeks gestation, the lower limit of viability, have the highest risks of all the complications of prematurity. Many biologic and environmental risk factors have been identified, and often overlap. For example, inflammation is associated with preterm birth and the development of BPD, white matter brain injury, necrotizing enterocolitis (NEC), and retinopathy of prematurity (ROP). How iNO exposure will influence the incidence of these complications of prematurity has been a major concern. Laboratory data suggest iNO may increase or decrease inflammation, cause bleeding by interfering with platelet aggregation and adhesion, and/or lead to accumulation of toxic substances (e.g., methemoglobin, formed by reaction of NO with hemoglobin, or nitrogen dioxide).

All 14 RCTs that compared treatment with iNO to standard treatment in preterm infants reported data regarding short term risks, including methemoglobin levels, and many complications of prematurity. The complications of prematurity we review in this section include brain injury, patent ductus arteriosus (PDA), sepsis, NEC, ROP, pulmonary hemorrhage, air leak, and pulmonary hypertension. Evidence of brain injury, obtained by serial head ultrasounds, includes intraventricular hemorrhage (IVH), intraparenchymal hemorrhage (IPH), hydrocephalus, periventricular leukomalacia (PVL), and other signs of white matter injury, including ventriculomegaly (Appendix E, Evidence Tables 3 and 4; Table 4). Meta-analyses were performed for all short term outcomes and are presented in a table at the end of this section. Not all RCTs reviewed in the text are included in the meta-analyses because of differences in the denominators across studies (e.g., all infants enrolled versus only survivors), and in the definition of the condition (e.g. any air leak versus only new air leak occurring after randomization). We grouped trials for analysis of each condition based on similar measurement characteristics, and indicate which trials were included in the table.

Table 4. Summary of outcomes for RCTs addressing KQ2.

Table 4

Summary of outcomes for RCTs addressing KQ2.

Evidence of brain injury. The nomenclature that describes injury to the preterm infant’s brain has changed since the publication of the earliest RCTs of iNO in preterm infants in 1997. Severity of IVH was often reported using the grading system proposed by Papile, 1978.87 Grade 1 is a germinal matrix hemorrhage (GMH), grade 2 is blood in the ventricle but not filling or dilating the ventricle, grade 3 is a large amount of blood in the ventricle and ventricular dilation, and grade 4 is blood in the brain parenchyma, i.e., intraparenchymal hemorrhage (IPH). In terms of its association with neurodevelopmental outcome, GMH is the most benign form of IVH, and despite the IVH grading system, it does not denote blood in the ventricle. In the very immature infant’s brain, the germinal matrix is a rich capillary network adjacent to the lateral ventricles, and very vulnerable to injury. Hemorrhage in the germinal matrix can extend into the ventricle, causing an intraventricular hemorrhage (IVH). The hemorrhage can also originate in the choroid plexus of the ventricle, and extend into the ventricle, causing an IVH. Blood in the ventricle may fill the ventricle and dilate it (Papile grade 3 IVH), or blood may be present in the ventricle with no ventricular dilation (Papile grade 2 IVH). However, there may be other causes of enlarged ventricles (called ventriculomegaly). Resorption of injured brain parenchyma can produce ventriculomegaly, as well as cysts in the brain parenchyma. Intraparenchymal hemorrhage (IPH) is more often caused by hemorrhagic infarction of brain tissue than by blood from an IVH extending into the brain parenchyma (Papile grade 4 IVH). An IPH may be due to blood filling the ventricles and compressing the venous network, or may be an injury to the brain that is unrelated to IVH. Periventricular leukomalacia (PVL) is seen when injured brain tissue, especially white matter, is resorbed and replaced by fluid. PVL manifests as small cysts, large cysts, larger ventricles with irregular borders, or any combination of these findings. PVL may be in the frontal, parietal or occipital lobes, and it may be on one side (unilateral) or bilateral. Some preterm infants who did not have an IVH develop ventriculomegaly due to resorption of injured brain. IVH with ventriculomegaly (Papile grade 3 IVH), IPH, PVL with or without ventriculomegaly are each associated with a high risk of neurodevelopmental impairment (NDI).87

Thirteen RCTs compared rate of brain injury on serial head ultrasounds in the iNO and control groups.34, 37, 39, 40, 58-62, 64-67 Brain injury may occur in utero, during labor and delivery, and immediately after birth, and is common in preterm infants on mechanical ventilation. Studies that compared head ultrasounds before treatment with head ultrasounds obtained after treatment can best determine whether exposure to iNO has a toxic or neuroprotective effect on brain injury. Few studies were able to obtain pretreatment head ultrasounds due to logistical problems. Only four RCTs obtained head ultrasounds at or before enrollment, and compared these to serial ultrasounds obtained during the remainder of the infant’s NICU hospitalization34, 37, 59, 64 (Appendix E, Evidence Table 8). The other seven RCTs did not obtain head ultrasounds before study entry.

Kinsella, 2006, enrolled 420 preterm infants born at and before 34 weeks gestation with a birth weight (BW) of 500 to 1250 g on mechanical ventilation within the first two days after birth, and treated them with low dose iNO (5 ppm) or placebo gas for 21 days.37 Head ultrasounds at study entry revealed no statistically significant differences between the iNO and placebo control groups in rates of GMH or IVH without ventriculomegaly (Papile grades 1 or 2 IVH, 18.4 percent versus 21.9 percent, respectively) or of IVH with ventriculomegaly (Papile Grade 3 IVH) or IPH (6.1 percent versus 6.6 percent, p-value = 0.41). Infants with GMH, IVH with or without ventriculomegaly or IPH (Papile grades 1 to 4 IVH) at study entry were reported in the outcome data if their condition worsened during or after treatment. Ultrasonographers were masked as to treatment category. They found no statistically significant differences of IVH with ventriculomegaly or IPH between iNO and placebo control groups, 12.3 percent versus 16.0 percent, RR 0.77 (0.54, 1.09) or of ventriculomegaly, 5.2 percent versus 8.9 percent, RR 0.58 (0.37, 1.01), p-value = 0.05. There was a statistically significant reduction of PVL in infants in the iNO group (5.2 percent) compared to placebo controls (9.0 percent), RR 0.58 (0.33, 1.00), p-value = 0.048. The infants in the iNO group had a statistically significant reduction in the rate of the composite variable of IVH with ventriculomegaly (Papile grade 3 IVH), IPH, PVL or ventriculomegaly than placebo controls, 17.5 percent versus 23.9 percent respectively, RR 0.73 (0.55, 0.98), p-value = 0.03 (Appendix E, Evidence Table 8).

Ballard, 2006 enrolled infants with a BW at or below 1250 g on ventilator support or CPAP at seven to 21 days, and treated them for a minimum of 21 days.34 Most preterm infants develop IVH or IPH within the first seven days after birth. Head ultrasounds were performed before enrollment and during and/or after administration of iNO or gas placebo. At baseline, there were no statistically significant differences in rate of unilateral IVH with ventriculomegaly or IPH, 11.9 percent versus 15.6 percent respectively; infants with bilateral IVH with ventriculomegaly or IPH were excluded. There were no differences between the iNO group and controls in the evolution of neurologic findings on head ultrasounds, 5.0 percent versus 4.1 percent, RR 1.21 (0.53, 2.76). (Appendix E, Evidence Table 8).

Two smaller RCTs were also able to perform head ultrasounds before initiating treatment. Kinsella, 1999 found that at study entry, 15 percent in the iNO group and 19 percent of controls had IVH with or without ventriculomegaly or IPH (Papile’s grades 2 to IVH).59 There were no statistically significant differences in rate of IVH or IPH with or without ventriculomegaly in survivors in the iNO group compared to controls, 28 percent versus 33 percent. They reported no statistically significant differences in rates of new or higher grade of IVH or IPH (44 percent versus 42 percent)59 (Appendix E, Evidence Table 9). In Subhedar, 1997, 42 infants born before 32 weeks gestation were enrolled at four days after birth and randomized to iNO or a control group.64 They obtained head ultrasounds at baseline and at weekly intervals for a month. No infant in either the iNO or the control group had an extension of an existing IVH64 (Appendix E, Evidence Table 8).

The small pilot RCT reported by Srisuparp, 2002 is the only RCT whose primary outcome variable was IVH with ventriculomegaly (Papile grade 3 IVH or IPH).66 They were unable to obtain head ultrasounds in all infants before study entry, however, as most were enrolled on the day of birth. They found no statistically significant differences between the iNO and control groups in brain injury, 25 percent versus 28 percent respectively.

Schreiber, 2003, enrolled 207 infants born before 34 weeks gestation with BW below 2000 g on mechanical ventilation for respiratory distress syndrome during the first week after birth.58 After randomization to the iNO or the gas placebo group, infants in the iNO group were given iNO at 10 ppm for 12 to 24 hours then 5 ppm for six days. They did not obtain head ultrasounds before study entry; all ultrasounds were interpreted by a pediatric radiologist masked to treatment assignments. Infants in the iNO group had statistically significantly lower rates of the composite variable, IVH with ventriculomegaly (Papile grade 3 IVH), IPH or PVL than placebo controls, 12.4 percent versus 23.5 percent respectively, (RR 0.53 (0.28, 0.98), p-value = 0.04). They found no statistically significant differences in the rate of posthemorrhagic hydrocephalus, 11.4 percent versus 9.8 percent, RR 1.17 (0.53, 2.58). (Appendix E, Evidence Table 8).

The secondary hypothesis of the Van Meurs 2005 RCT of infants born before 34 weeks gestation with BW 401 to 1500 g who had severe respiratory failure was that iNO would not increase the incidence of the composite variable, IVH with ventriculomegaly (Papile grade 3 IVH), IPH or PVL.40 This study was terminated after the second planned analysis because of a higher rate of the composite brain injury variable in the iNO group than in controls reached statistical significance. However, when outcomes were analyzed for all 420 enrolled infants (the plan was for 440 infants) there were no statistically significant differences in rates of the composite brain injury variable (Papile grade 3 IVH, IPH or PVL) whether ultrasounds were read by each center’s local radiologists, 39 percent in the iNO group and 32 percent in controls, RR 1.25 (0.95, 1.66); or when they were read by a central masked reader after the study was terminated, 37 percent versus 38 percent, RR 0.97 (0.74, 1.27). Infants enrolled in this RCT had similar BW as in Kinsella 2006, and both RCTs had lower BW than in Schreiber, 2003.37 58 However, infants in Van Meurs, 2005 were sicker than those in either Schreiber, 2003 or Kinsella, 2006, with OI 22 to 23 compared to five to seven at enrollment (Appendix E, Evidence Table 8).

In a meta-analysis of five RCTs34, 37, 39, 40, 58 that reported the composite brain injury variable, defined by a combination of IVH with ventriculomegaly, IPH, or PVL (Kinsella, 2006 included ventriculomegaly as a separate variable), there was no statistically significant difference between infants treated with iNO and controls, RR 0.86 (0.58, 1.29). Results were unaffected by removal of the Ballard trial,34 a study that enrolled infants much later than the other trials included in the analysis and reported only new or worsening brain injury: RR 0.79 (0.50, 1.27) (Figure 9). There was a substantial degree of heterogeneity among the five studies in this meta-analysis of brain injury (I2 = 0.657). The two RCTs with the lowest RR of brain injury (Van Meurs, 2007 and Schreiber, 2003) differed from the other studies by including larger preterm infants, with BW above 1500 g.39, 58 Brain injury tends to occur during the first week after birth and is associated with cardiovascular instability in sick preterm infants. We can speculate that the larger preterm infants derived greater benefit from the effect of iNO on cardiovascular stability, as is seen with more mature full term infants. Smaller, more preterm infants may not benefit as much from this effect, due to immature autoregulation of their cerebral blood flow.

This figure is a forest plot of five RCTs used to calculate whether pooled data on the impact of iNO on brain injury, in preterm infants, is significant

Figure 9

Meta-analysis of five studies describing brain injury.

Similarly, a meta-analysis of RCTs that reported the incidence of PVL showed no difference between the iNO and control groups, RR 0.78 (0.37, 1.62) (Table 5).

Table 5. Meta-analyses of short term risks of iNO therapy.

Table 5

Meta-analyses of short term risks of iNO therapy.

In summary, one large multicenter RCT and one large single center RCT found a lower rate of brain injury (IVH with ventriculomegaly, IPH, PVL, +/- ventriculomegaly) in infants treated during the first week after birth with iNO compared to placebo controls. Another large multicenter RCT was terminated early for concern that the iNO group had a higher rate of IVH with ventriculomegaly, IPH or PVL than controls, but on final analyses, there were no statistically significant differences between the iNO and control groups. All the other RCTs found no statistically significant differences between the iNO and control groups in rates of all IVH (Papile grades 1 to 4), IVH with ventriculomegaly IPH, PVL, hydrocephalus, or combinations of these variables. What makes these findings important is that these signs of brain injury on serial head ultrasounds in the NICU are some of the best predictors for neurodevelopmental impairment in preterm infants. Key Question 3 addresses more long term outcomes, at a year or more, including cerebral palsy (CP), cognitive abilities, and neurodevelopmental impairments.

Patent ductus arteriosus. In the fetus, the ductus arteriosus allows most of the blood to bypass the lungs (and circulate through the placenta). In preterm infants, especially the most immature, failure of this duct to close can interfere with their transition to extrauterine life and lead to heart failure. By altering pulmonary blood flow, iNO may influence duct closure. Eleven RCTs described in 12 articles compared incidence of PDA in the iNO group and controls.34, 37, 58, 59, 61-67, 78 Some trials reported only those infants who underwent surgical ligation of their PDA, and others included all infants diagnosed with PDA, whether they were treated medically or surgically. Kinsella, 2006 reported rates of symptomatic PDA that were medically treated (54.0 percent in the iNO group versus 53.7 percent of controls), and rates of PDA treated with surgical ligation (21.6 percent versus 21.8 percent).37 None of the eleven RCTs (Appendix E, Evidence Table 9) or a meta-analysis (RR 1.01 (0.86, 1.19); Table 5) found a statistically significant difference in incidence of PDA between the iNO groups or controls.

Sepsis. Eight RCTs reported data on infants who developed sepsis. Schreiber, 2003 reported the incidence of sepsis diagnosed after the first day, to distinguish between infants who were septic at birth from those that developed sepsis during their NICU course.58 Some studies reported sepsis only if the infant’s blood culture was positive.34, 63, 66, 67 None of the eight RCTs34, 37, 58, 62, 63, 65-67 that reported rate of sepsis found statistically significant differences between their iNO and control groups (Appendix E, Evidence Table 9). All eight trials were included in a meta-analysis that found no difference in the development of sepsis between infants treated with iNO and controls, RR 1.06 (0.95, 1.18) (Table 5).

Necrotizing enterocolitis. NEC is an acute inflammation of the intestines that can lead to intestinal perforation, surgical resection of injured bowel and placement of an ostomy. Bowel perforation is generally associated with sepsis, and treatment consists of intravenous antibiotics, bowel rest, parenteral nutrition, and cautious refeeding. NEC can therefore have an impact on subsequent health and growth. Eight RCTs reported in nine articles34,37, 58, 61, 62, 64, 65, 67,78 compared the incidence of NEC in iNO and control groups. Ballard, 2006 was the only study to distinguish between NEC treated medically and infants who needed surgery. The They found no statistically significant differences in incidence of NEC, 7.8 percent in the iNO group versus 6.6 percent in controls, RR 1.17 (0.64, 2.13) or NEC requiring surgery (3.4 percent in the iNO group and 2.8 percent in controls, RR1.20 (0.46, 3.13).34 None of the eight RCTs (Appendix E, Evidence Table 9) nor our meta-analysis (RR 1.23 (0.94, 1.62; Table 5)) found any statistically significant differences in NEC between iNO and control groups.

Retinopathy of prematurity. Retinopathy of prematurity is a neovascular retinal disorder, which can result in severe visual impairment. Serial eye examinations determine whether ROP is present as the retina is vascularized, and if it is progressing. Visual outcomes are improved for severe ROP, especially if there are dilated, tortuous blood vessels in the posterior pole of the eye (i.e. plus disease) with laser surgery. Eight RCTs report the incidence of severe ROP treated with laser surgery.34, 37, 39, 40, 58, 59, 63, 64 Ballard, 2006 found a high incidence of any degree of ROP in their high risk study population, 83.7 percent in the iNO group and 81.9 percent in controls, RR 1.00 (0.93, 1.07).34 Their incidence of severe ROP requiring treatment was 24.5 percent in the iNO group versus 23.6 percent in controls, RR 0.97 (0.72, 1.31). This is similar to the incidence of ROP requiring treatment in the other seven RCTs, and none found statistically significant differences between iNO and control groups (Appendix E, Evidence Table 9). A meta-analysis confirmed no statistically significant difference in ROP between infants treated with iNO and controls, RR 1.01 (0.82, 1.24; Table 5).

Pulmonary complications. In Key Question 1, we addressed the primary pulmonary complication of prematurity, BPD. In this section, we report other pulmonary complications: pulmonary hemorrhage, air leak or pneumothorax, pulmonary hypertension or right heart failure. An important consideration is whether infants were excluded from studies if they had evidence of bleeding or air leak before entry into the study. If they were not excluded, the most meaningful data are the rate of pulmonary hemorrhage or air leak once entered into the study. Five RCTs, described in six articles, excluded infants with low platelets or bleeding problems,40, 61, 63, 65, 67, 62 and four excluded infants with severe intracranial or pulmonary hemorrhage.34, 37, 60

Seven RCTs report data on pulmonary hemorrhage. Whether they excluded infants with bleeding problems37, 63, 65 ,74 or not58, 59, 64 they did not find any statistically significant differences between iNO and control groups in rates of pulmonary hemorrhage (Appendix E, Evidence Table 9). Our meta-analysis with trials that excluded infants with bleeding problems showed no difference in pulmonary hemorrhage between iNO treated infants and controls, RR 0.89 (0.60, 1.33) (Table 5).

Ten RCTs reported rates of air leak or pneumothorax, and none found any statistically significant differences between the iNO and control groups.37, 39, 40, 58, 59, 62-66 Schreiber, 2003 reported pneumothorax and pulmonary interstitial emphysema separately, finding no statistically significant differences in rate of pneumothorax (10.5 percent versus 16 percent, respectively) or pulmonary interstitial emphysema (27 percent versus 34 percent, respectively).58 The rates of air leak varied from a low of four to six percent58, 64, 65 to as high as 35 to 38 percent40, 63 (Appendix E, Evidence Table 9). Our meta-analysis with trials that included all infants in the denominator also found no difference in the risk of air leak between the iNO treated infants and controls, RR 0.96 (0.71, 1.28) (Table 5).

The only trial that reported pulmonary hypertension as an outcome variable documented 50 percent of infants in the iNO and control group with the condition.67 No study specifically documented right heart failure (Appendix E, Evidence Table 8).

Methemoglobinemia. Twelve RCTs measured methemoglobin levels, and some measured nitrogen dioxide levels in administered gas.34, 37, 39, 40, 58, 59, 62-67 Most reported that methemoglobin levels in all infants were not elevated,34, 59, 66, 67 or were below 2.5 percent,65 three percent,64 or four percent.39 The Van Meurs, 2005 RCT of infants born before 34 weeks gestation with BW 400 to 1500 g found two infants (1 percent) in each group who had methemoglobin levels above four percent.40 One infant in the iNO group had a methemoglobin level of at least eight percent, and the nitrogen dioxide level was at or above 3 ppm in two percent, and at or above 5 ppm in one percent. The multicenter Kinsella, 2006 trial reported a transient mild elevation of methemoglobin level in two of 398 (0.05 percent) infants, but elevation was not defined.37 Three infants treated with iNO in the Schreiber, 2003 RCT had elevation in methemoglobin level that never rose above seven percent, and nitrogen dioxide was never above 2 ppm.58 The Field RCT allowed the highest maximum dose of iNO, up to 40 ppm, and as many as eight of 55 (14.5 percent) preterm infants had methemoglobin levels above two percent; only one infant (1.8 percent) had nitrogen dioxide above 2 ppm for 30 minutes63 (Appendix E, Evidence Table 9).

Conclusion

Key Question 2 analyzed 14 RCTs of iNO in preterm infants on mechanical ventilation for evidence of toxicity or short term risks of iNO. None of the 14 RCTs reported statistically significant effects of iNO on rates of PDA, sepsis, NEC, treated ROP, pulmonary hemorrhage, or air leaks. No study reported toxic accumulations of methemoglobin. None of the 13 RCTs that reported head ultrasound evidence of brain injury reported a statistically significant increase with iNO treatment. Two large RCTs, with more than 100 subjects in each group, reported a statistically significant reduction of a composite brain injury variable (IVH with ventriculomegaly, IPH or PVL) in the iNO group compared with placebo gas controls.37, 58 These two RCTs raise the question as to whether iNO has neuroprotective effects. There was no statistically significant difference between the iNO and control groups in a meta-analysis that pooled data from five RCTs that reported rates of the composite brain injury variable (IVH with ventriculomegaly, IPH or PVL). There was also no statistically significant difference in our meta-analysis of four RCTs with data on rates of PVL. However, not only do the RCTs vary widely in study design, but there is also little uniformity among studies as to when head ultrasounds were performed, who interpreted them (locally at each center or at more uniformly at one site), categories reported, and criteria used for each category. These RCTs were generally powered for death and BPD, and not for short term risks or brain injury. There is insufficient evidence for assessing the effect of iNO on the preterm infant’s brain. There is a need for RCTs that obtain neuroimaging before initiation of treatment and at regular prespecified intervals, provide for uniform interpretation of neuroimaging studies, carefully define categories of types of brain injury, and clearly report rates of each type, and composites of brain injury in terms of surviving infants. Because they are so vulnerable as they are rapidly maturing, the effects of any intervention on the brain should be studied in every RCT involving preterm infants. Key Question 3 reviews the evidence of effects of iNO on longer term neurodevelopmental, pulmonary, and other health outcomes.

Key Question 3. Are there effects of iNO therapy on long term pulmonary and/or neurodevelopmental outcomes among premature infants who receive respiratory support?

Major Findings

  • There is insufficient evidence to determine whether iNO therapy in preterm infants who require respiratory support influences the incidence of cognitive, motor or sensory impairments, or neurodevelopmental disability.
  • There is evidence suggesting that iNO therapy in preterm infants who require respiratory support may decrease the use of respiratory medications at one year of age.
  • There is insufficient evidence to determine whether iNO therapy in preterm infants who require respiratory support impacts long term health outcomes such as lung growth and development, pulmonary morbidity, rehospitalization after NICU discharge, and growth.

Detailed Analysis

Nine articles representing six RCTs report long term followup of health and neurodevelopmental outcomes at one year corrected for degree of prematurity or later (see Table 6). Field, 2005 reported on some health and neurodevelopmental outcomes at one year corrected for degree of prematurity of the multicenter INNOVO RCT.63 Mestan, 2005 reported neurodevelopmental outcomes and growth at two years of the infants enrolled in Schreiber, 2003, the largest single center RCT.56 Hintz, 200758 reported on survival, CP, cognitive abilities and neurodevelopmental impairment (NDI) in 18 to 22 month old survivors enrolled in the NICHD RCT of infants born before 34 weeks gestation with birth weight below 1500 g.30, 40 Neurodevelopmental impairment at one year corrected for degree of prematurity is included in the Van Meurs, 2007 paper that reported results from the NICHD RCT on infants born before 34 weeks gestation with birth weight above 1500 g.39 For surviving infants in Ballard, 2006, Walsh, 2010 reported on neurodevelopmental outcomes and growth at two years of age, corrected for degree of prematurity, and Hibbs, 2008 reported on pulmonary and health outcomes at one year.34, 44, 57 In a paper focused mostly on economic costs and resource utilization, Watson, 2009 reported on survival and some neurodevelopmental outcomes at one year of age, corrected for degree of prematurity, for infants enrolled in Kinsella, 2006.36, 37 Bennett, 2001 reported on 30 month survival for all study participants who were discharged from the NICU, and neurodevelopmental outcomes for 21 of the 22 children alive at 30 months, corrected for degree of prematurity.76 Huddy, 2008 followed the group of infants in Field, 2005 up to four to five years, and reported on several health and neurodevelopment related outcomes; this is the longest followup for any of the RCTs35 (Appendix E, Evidence Tables 3 and 4; Table 6).

Table 6. Summary of outcomes for RCTs addressing KQ3.

Table 6

Summary of outcomes for RCTs addressing KQ3.

Trials that reported comparable neurodevelopmental outcomes were included in meta‐analyses. There was some variability in the incidence of outcomes among the few trials that reported conditions such as CP, vision, and hearing impairment. The variability is likely due to the low prevalence of these conditions and small samples, as studies were not powered to detect difference in these outcomes. Few trials reported other long term health outcomes in a consistent manner, making pooled estimates of risk impossible, with the exception of pulmonary outcomes.

Death and survival beyond the NICU. Followup studies of two RCTs reported survival into early childhood. Huddy, 200835 followed children from Field, 200563 until four to five years of age. A total of 108 infants were enrolled in the RCT, 44 survived to their first birthday. Overall survival to four to five years was 44 percent in the iNO group and 36 percent in controls. Mestan, 200556 reported that 85 percent of the iNO group and 77 percent of placebo controls from Schreiber, 200358 were alive at two years. Additionally, seven followup studies reported long term mortality rates for six RCTs. Study results are displayed in Table 7. None of the studies revealed a significant difference in mortality when comparing infants treated with iNO to controls. (Appendix E, Evidence Table 10).

Table 7. Studies addressing death and/or survival beyond the NICU.

Table 7

Studies addressing death and/or survival beyond the NICU.

A meta-analysis was conducted with all the trials that reported death at any time after NICU discharge, regardless of the age of the children at the time of the measurement. Two studies were excluded (Hibbs44 and Huddy35) because there was more than one followup study for the Ballard and Field trials. The pooled estimate shows no difference in mortality with iNO therapy compared to placebo, RR 1.02 (0.86, 1.20) (Figure 10).

This figure is a forest plot of seven RCTs used to calculate whether pooled data on the impact of iNO on death after NICU discharge, in preterm infants, is significant

Figure 10

Meta-analysis of death at followup after NICU discharge.

Cerebral palsy and motor outcomes. Cerebral palsy is a disorder of movement and posture caused by malformation or injury to the developing brain that cannot be diagnosed in the neonate, but requires a neurological examination and assessment of motor function at one or more years after birth. Cerebral palsy varies in terms of type (spasticity, extrapyramidal or mixed), anatomic distribution (diplegia, hemiplegia, etc.) severity, and associated disabilities (cognitive and/or sensory impairments). The more severe the CP, the earlier it can be diagnosed; diagnosis of mild CP is generally not made until two years or more. Diagnosis of CP requires a comprehensive neurodevelopmental examination focusing on abnormalities of muscle tone, deep tendon and other reflexes, movement and posture, as well as an assessment of motor function. The most common type of CP in preterm infants is spastic diplegia, which involves increased muscle tone and reflexes in both lower extremities with little or no involvement of the upper extremities. CP prevalence increases with decreasing gestational age and birth weight. Most studies reported moderate to severe CP. The functional classification for CP is included in the description of each study that reported this outcome.

The Hintz, 200730 18 to 22 month followup study of Van Meurs, 200540 RCT of infants with birth weight 400 to 1500 g found CP in 20 percent of the iNO group and 11 percent of controls. CP functional impairment was defined as the ability to sit independently or with support but not ambulate independently (moderate CP), or the inability to sit or walk without support (severe CP). The initial RR was not significant at 1.85 (0.93, 3.71). When adjusted for birth weight, OI entry criterion strata, sex, BPD, IVH 3 or 4 or PVL, postnatal steroid exposure, study center, and length of iNO exposure, the RR was significant at 2.41, indicating a higher rate of CP in iNO treated infants, but with a wide 95 percent confidence interval (1.01, 5.75) (Appendix E, Evidence Table 11).

Tanaka, 200738 evaluated a cohort of children at three years of age who had received iNO or 100 percent oxygen in the neonatal period for hypoxic respiratory failure with pulmonary hypertension. Cerebral palsy, defined as abnormal muscle tone in one extremity and abnormal control of movement and posture, was diagnosed in 12.5 percent of those treated with iNO compared to 46.7 percent who had been treated with 100 percent oxygen (p-value = 0.054) (Appendix E, Evidence Table 11). There was also a significantly lower odds of CP in children who had received iNO versus 100 percent oxygen (OR=0.16; 0.03, 0.98). This association persisted in several multivariate models.

The other five RCTs that evaluated for CP found no significant differences in the incidence of CP in the iNO group compared to controls. The Van Meurs, 2007 RCT of infants with birth weight above 1500 g found that none of the 17 infants who were followed to one year corrected for degree of prematurity developed CP.39 In the Mestan, 2005 paper that reported two year outcomes of survivors of the Schreiber, 2003 RCT, CP rates were virtually the same, nine percent in the iNO group and 10 percent in controls.56 They based their diagnoses of CP and its type on abnormalities in neuromotor tone, deep tendon reflexes, primitive reflexes, postural reactions, movement or coordination, and delay in motor milestones.58 Walsh, 2010 reported similar findings from the Ballard, 2006 cohort: six percent of iNO treated infants and five percent of control infants developed CP by two years.34, 57 Motor functional impairment for CP was determined by Palisano’s Gross Motor Function Classification Scale (at or above 2).88 Of the seven infants in the iNO group of Subhedar, 1997 that were followed to 30 months none developed CP, compared to two of 14 controls (14 percent), who had significant abnormalities of tone or movement.64, 76 In the Huddy, 2008 report of four to five year outcomes of Field 2005, the CP rate (moderate to severe disability of motor function) was 13.6 percent in the iNO group and 12.5 percent in controls.35, 63

A meta-analysis of the seven trials that evaluated motor outcome found no statistically significant difference in CP among infants treated with iNO compared with controls, RR 1.07 (0.67, 1.71) (Figure 11). A separate meta-analysis was performed with four trials that used the Bayley Scales Psychomotor Developmental Index below 70 to define motor delay.30, 39, 56, 76 Individually, none of these trials found a statistically significant difference in the incidence of motor delay when comparing those who had received iNO to controls. Similarly, the meta-analysis showed no statistically significant difference in the incidence of a motor delay with iNO therapy, compared with controls, RR 0.95 (0.66, 1.36) (Figure 11).

This figure is a forest plot of seven RCTs used to calculate whether pooled data on the impact of iNO on cerebral palsy, in preterm infants, is significant

Figure 11

Meta-analysis of cerebral palsy.

Cognitive outcomes. There were six RCTs and one cohort study that reported cognitive outcomes. The majority used the Bayley Scales of Infant Development Mental Developmental Index (MDI) for assessment and defined cognitive impairment as MDI < 70, two standard deviations below the mean. The only followup study to report a statistically significant difference in cognitive impairment between the iNO group and controls was the followup of Schreiber, 2003 reported by Mestan, 2005.56, 58 Their followup rate was 82 percent at two years corrected for degree of prematurity. They found that only 19 percent of the iNO group had a Bayley MDI score more than two standard deviations below the mean compared to 35 percent of controls, p-value = 0.03. This result must be considered in the context of the significantly lower rate of the combined variable of grade 3 IVH, IPH, and PVL in the iNO group compared to controls as reported in Schreiber, 2003 (Appendix E, Evidence Table 12).

Hintz, 2007 evaluated participants with birth weight 400 to 1500 g enrolled in Van Meurs, 2005 at 18 to 22 months of age. Forty-three percent of infants in the iNO group had MDI scores more than two standard deviations below the mean compared to 36 percent of controls, RR 1.2 (0.84, 1.73).30 Infants in the Van Meurs, 2007 RCT with birth weight above 1500 g were followed to one year corrected for degree of prematurity. In the iNO group, 11 percent had MDI scores more than two standard deviations below the mean, compared to 25 percent of controls, RR 0.44 (0.50, 4.02).39

A meta-analysis was performed using these three studies in which cognitive impairment was defined as MDI < 70. This revealed no statistically significantly difference between those treated with iNO therapy and controls, RR 0.78 (0.39, 1.60) (Figure 12). As in the meta-analysis for the brain injury, there is substantial heterogeneity, reflecting that many of the same infants are included in this meta-analysis. Again, the Van Meurs, 200739 and Mestan, 200556 (followup of Schreiber, 2003) studies included infants with birth weight above 1500 g with a lower risk for brain injury and subsequent cognitive impairment than the Hintz30 study (followup of Van Meurs, 2005) that restricted enrollment to those with birth weight of 400 to 1500 g.

This figure is a forest plot of four RCTs used to calculate whether pooled data on the impact of iNO on MDI < 70, in preterm infants, is significant

Figure 12

Meta-analysis of cognitive development as measured by the Bayley Scales Mental Developmental Index below 70.

In their two year followup of Ballard, 2006, Walsh, 2010 reported cognitive outcomes in terms of normal intelligence, defined as MDI score above 85, one standard deviation below the mean.34, 57 There was no significant difference in proportion of survivors with MDI above 85; there were 39 percent in the iNO group and 35 percent in the placebo control group. Translating these data into the proportion with cognitive delay, 61 percent in the iNO group and 65 percent in the placebo control group had MDI scores one standard deviation below the mean or lower. They also reported mean MDI scores for each group and found no significant difference: 81 +/- 20 versus 79 +/- 22 (Appendix E, Evidence Table 12). Bennett, 2001 reported the incidence of cognitive delay, defined as MDI < 85 in survivors from Subhedar, 1997 at 30 months of age corrected for prematurity.76 There was no significant difference in the incidence of cognitive neurodevelopmental delay when comparing those treated with iNO to controls, RR 0.89 (0.37, 1.75).

To evaluate cognition at four to five years, the Huddy, 200835 followup of the Field, 2005 cohort used the British Ability Scales (BAS), 89 which has norms similar to the Bayley and other intelligence tests, with a standardized mean of 100 and a standard deviation value of 15.35 Three children in the iNO group and one control had severe impairments that precluded using the BAS. There were no statistically significant differences in mean General Conceptual Ability Score (GCAS) between the 19 children in the iNO group and the 15 children in the control group: 91.2 +/- 21.1 versus 81.3 +/- 22.5. They also found no statistically significant differences in the BAS cluster scores for verbal ability, pictorial reasoning, spatial abilities, and the nonverbal composite scores. There were six of 22 children (27 percent) in the iNO group with GCAS scores two or more standard deviations below the mean, compared with six of 16 controls (38 percent) (Appendix E, Evidence Table 12).

Sensory impairment. There were no significant differences between the iNO and control groups in proportion of children with visual impairment or hearing impairment in seven studies (representing six original trials) that report these outcomes. Visual impairments occurred in zero to four percent of children in the iNO group compared to zero to four percent in controls in the six studies that reported this outcome.63 Our meta-analysis that included trials reporting early childhood blindness revealed no significant difference between those treated with iNO therapy and controls, RR 1.09 (0.52, 2.30) (Figure 13). Hearing impairments occurred in zero to nine percent of children in the iNO group compared to zero to seven percent of controls in the same six followup studies30, 39, 56, 57, 63, 76 (Appendix E, Evidence Table 13). The pooled risk ratio for hearing loss also showed no significant difference with iNO therapy compared to controls, RR 1.50 (0.69, 3.27) (Figure 14).

This figure is a forest plot of six RCTs used to calculate whether pooled data on the impact of iNO on visual impairment, in preterm infants, is significant.

Figure 13

Meta-analysis of visual impairment.

This figure is a forest plot of six RCTs used to calculate whether pooled data on the impact of iNO on visual impairment, in preterm infants, is significant.

Figure 14

Meta-analysis of hearing impairment.

Neurodevelopmental impairment. Seven studies reported the proportion of children with neurodevelopmental impairment (NDI), a combined variable that included cognitive, neuromotor, and sensory impairments. Children with moderate to severe CP were included, as were children with severe visual or hearing impairments. All studies defined “cognitive impairment” as two or more standard deviations below the mean score for the assessment tool that was used. Most studies also included children with Psychomotor Developmental Index scores two or more standard deviations below the mean from the Bayley Scales of Infant Development30, 36, 39, 56, 57, 76 (Appendix E, Evidence Table 14, Table 8).

Table 8. Studies addressing neurodevelopmental impairment.

Table 8

Studies addressing neurodevelopmental impairment.

Just as they found statistically significant differences in cognitive impairment, the Mestan, 2005 two year followup study of Schreiber, 2003 found that NDI rates were lower in the iNO group compared to placebo controls: 24 percent versus 46 percent, RR 0.53 (0.33, 0.87), p-value = 0.01. The six other followup studies revealed no significant differences between the two groups.35, 39, 56-58, 76 In two large RCT two year followup studies, as many as half of the survivors in both the iNO and control groups had NDI: 45 percent versus 49 percent respectively, RR 0.92 (0.75, 1.12) reported by Walsh, 2010 using Ballard, 2006 cohort; and 51 percent versus 47 percent respectively, RR 1.07 (0.8, 1.44) reported by Hintz, 2007 for the Van Meurs, 2005 multicenter RCT.57, 30 The Van Meurs, 2007 RCT of infants with birth weight above 1500 g reported a lower rate of NDI in both iNO and control groups, with no statistically significant difference between the two groups: 11 percent versus 25 percent, RR 0.44 (0.5, 4.02).30, 39 Despite the finding in Kinsella, 2006 of a lower rate of grade 3 IVH, IPH or PVL in infants in the iNO group, Watson, 2009 reported no statistically significant differences in the rate of NDI at one year corrected for degree of prematurity in infants in the iNO group compared to controls, 35 percent versus 34 percent.36, 37 Huddy, 2008 reported no significant differences in four to five year old children from the Field, 2005 cohort, with NDI in 36 percent of children in the iNO group and in 44 percent of controls.35, 63 Bennett, 2001 reported that for 30 month old children in the Subhedar, 1997 RCT, none of the seven survivors had NDI compared to 36 percent of controls64, 76 (Appendix E, Evidence Table 14).

Our meta-analysis of trials that measured outcome at 12 to 30 months suggests no statistically significant difference in the proportion of infants with NDI between those given iNO versus the control group (RR 0.91 (0.77, 1.12)) (Figure 15).

This figure is a forest plot of six RCTs used to calculate whether pooled data on the impact of iNO on NDI, in preterm infants, is significant.

Figure 15

Meta-analysis of studies reporting NDI.

Two followup studies reported the rate of children in each group who had no impairment. For the followup of infants from Kinsella, 2006, Watson, 2009 defined “no impairment” to include only those children who had MDI and Bayley Physical Developmental Index (PDI) above 85, and no CP or severe visual or hearing impairment.36, 37 They found no statistically significant differences in the proportion of children with no impairments at one year corrected for degree of prematurity between the iNO group and controls: 38 percent versus 37 percent. In reporting the 18 to 22 month followup results from Van Meurs, 2005 RCT on preterm infants with birth weight 400 to 1500 g, Hintz, 2007 used a similar definition of “unimpaired”: MDI and PDI ≥ 85, no moderate to severe CP and not blind or deaf.30 They found that 23 percent in the iNO group and 25 percent in placebo controls were unimpaired. The low proportion of survivors with no impairments is an indication of how sick the infants enrolled in the RCTs were. Conversely, there was a higher survival rate among infants with NDI. (Appendix E, Evidence Table 13).

Death or neurodevelopmental impairment. None of the four studies that reported the rate of the composite variable, death or NDI, for infants enrolled in four RCTs found any significant differences between the iNO and control groups.30, 36, 39, 76 Although Kinsella, 2006 reported lower rates of grade 3 IVH, IPH or PVL in infants in the iNO group compared to controls, Watson, 2009 reported no significant differences in the rate of death or NDI at one year corrected for degree of prematurity: 42.4 percent in the iNO group and 44.5 percent in placebo controls.36 Similarly, in the two Van Meurs RCTs, there were no significant differences between the iNO groups and placebo controls; Hintz, 2007 reported that 78 percent of preterm infants with birth weight 400 to 1500 g in the iNO group died or had NDI compared to 73 percent of controls, RR 1.07 (0.95, 1.19)30; while Van Meurs, 2007 reported that 43 percent of preterm infants with birth weight above 1500 g in the iNO group died or had NDI compared to 50 percent of placebo controls, RR 0.86 (0.37, 1.96).39 In the Bennett, 2001 followup of Subhedar, 1997, 63 percent in the iNO group and 59 percent in the control group died or had NDI at 30 months 64, 76 (Appendix E, Evidence Table 14).

Long term health outcomes.

Seizures. Seizures can accompany complications that occur in the antenatal period or in the NICU, including perinatal asphyxia, hypoxia, hypoglycemia and other electrolyte abnormalities, intraventricular hemorrhage and meningitis. Seizures in premature infants that persist beyond the NICU are usually a result of brain injury and associated with other neurodevelopmental sequelae. Field, 200563 included “on anticonvulsants” and “fits in previous four weeks” with neuromotor outcomes at one year corrected age. These outcomes are based on available pediatrician assessments. In the iNO group, three of the 25 infants, and one of the 18 control infants with available reports were being treated with anticonvulsants. Three infants in the iNO group and none of the 18 control infants had experienced a fit or seizure in the four weeks prior to the assessment. Seizures were included in the General Health domain of the four to five year assessments of infants enrolled in the Field, 2005 RCT performed by Huddy, 2008.35 Children could be categorized as normal, impaired, or mildly, moderately, or severely disabled with regard to seizures. Of the five children who had seizures in the 12 months prior to assessment (3 of 22 iNO, 2 of 16 controls), three iNO and one control were on regular seizure medications, and considered to be impaired. One iNO child had more than one seizure per month and was classified as mildly disabled (Appendix E, Evidence Table 15).

Growth. There are five RCTs and one cohort study in which growth parameters were included with early childhood outcomes. Cheung, 1998 evaluated the 10 survivors from a cohort of 24 infants who received rescue iNO therapy for severe hypoxemic respiratory failure. One or more anthropometric measures (weight, length, head circumference) of four of the 10 (40 percent) infants evaluated at 13 to 40 months were below the third percentile on a standard growth curve when plotted at their corrected age.72 In the followup of survivors from the Van Meurs, 2005 RCT at 18 to 22 months,30 Hintz, 2007 measured weight and head circumference. When comparing the infants who had received iNO to controls, there was no difference in the measures of weight and head circumference, or percentage of infants with weight or head circumference below the fifth percentile for corrected age, based on CDC growth charts90 (Appendix E, Evidence Table 15).

Infants enrolled in the Field, 2005 trial had growth parameters reported at one year corrected age,63 and four to five years chronologic age.35 At one year, there were pediatrician reports on 25 infants who had received iNO, and 18 controls.63 Infants were categorized by the number of standard deviations their length and weight were from a standardized height and weight.91 There was no reported analysis to determine whether this distribution was different between the two groups. The actual measure of head circumference was reported. The mean head circumference was similar in the two groups, within one standard deviation: iNO 45.5 cm (1.8), control 45.2 cm (1.6). In the followup study by Huddy, 2008, weight, length, and head circumference were measured in 22 participants from the iNO group and 15 controls at four to five year of age.35 There were no differences in the standardized mean values91 for any of the three parameters between the iNO and control groups. It was noted that values were lower in both groups than those of normal population. Walsh, 201057 and Mestan, 200556 both evaluated infants at two years of age corrected for prematurity from Ballard 2006 multicenter RCT, and Schreiber 2003 single center RCT, respectively. In the former, there were no significant differences in measures of weight, length or head circumference. Mestan, 200556 reported the measures and generated z-scores using the CDC growth charts,90 which revealed that both the iNO and control groups were smaller than the reference population for all measures. Unlike the others who reported growth measures, Mestan, 200556 found that those in the iNO group were significantly heavier than participants in the control group (median weights 11.7 kg versus 10.8 kg, p-value = 0.04; z-scores -0.49 versus -1.07, p-value = 0.02); and measures of length and head circumference were not different (Appendix E, Evidence Table 15). Participants lost to followup had a higher birth weight and greater gestational age at delivery, which could influence followup weight if they were not equally distributed among the iNO and control groups.

Oral feeding. Successful oral feeding requires the coordination of basic reflexes, more complicated motor skills, and effective breathing. Infants must coordinate these efficiently in order to take in enough to support the energy expenditure required for this task, as well as growth. Any of the required skills can be adversely affected by premature birth and associated complications. Lung disease can increase the required energy expenditure that is necessary for maintenance and catch up growth. Only Field, 200563 reported oral feeding and did so as a secondary outcome measure. Reports from pediatricians revealed that three of 25 infants who received iNO had a stoma for feeding. Of these infants, only one was limited to feeding through the stoma only; two infants also took some pureed feeds orally. One infant who did not receive iNO (of 18) was limited to liquid feeds through a tube but did not have a surgical ostomy placed for feeding. Five iNO infants and four controls could only manage pureed foods. The remainder (17 iNO and 13 control infants) could also manage eating lumps (Appendix E, Evidence Table 15).

Pulmonary and other health outcomes. There were two cohort studies and four randomized controlled trials that reported pulmonary outcomes beyond NICU hospitalization. The reported markers of pulmonary health varied among studies and included the use of supplemental oxygen or respiratory medications, asthma or wheezing, respiratory disability, feeding tube, and recurrent aspiration.

The safety and efficacy study by Clark, 200271 included infants ≤ 1250 grams birth weight, on mean airway pressure of ≥ 7cm H20 and FiO2 ≥ 40 percent at 10 to 30 days of age. The focus of the study was safety and short term efficacy. However, records of 25 of the 29 survivors were available at six months of age and revealed that 10 of the infants continued to require supplemental oxygen (40 percent) (Appendix E, Evidence Table 15).

Cheung, 1998 reported on the ten survivors from the cohort of 24 infants who received iNO as rescue therapy for severe hypoxemic respiratory failure. Eight were diagnosed with bronchopulmonary dysplasia; all had supplemental oxygen discontinued by 10 months corrected age. Other pulmonary issues reported at followup that occurred in the range of 13 to 40 months corrected age include recurrent aspiration pneumonia (1/10), and chronic lung disease requiring bronchodilator therapy on a regular basis (1/10). Four of the 10 children had recurrent wheezing episodes and used bronchodilator therapy intermittently72 (Appendix E, Evidence Table 15).

Hibbs, 2007 reported the pulmonary outcomes at one year of 85 percent of the survivors enrolled in Ballard, 2006.34, 44 The control group had a greater prevalence of reported pulmonary morbidity at one year of age when compared to the iNO group, based on respiratory symptoms (56.4 percent versus 49.6 percent ; OR 0.70 (0.48–1.03)), use of diuretics (28.4 percent versus 18.6 percent; OR 0.54 (0.34, 0.85)), systemic (17.7 percent versus 11 percent; OR 0.56 (0.32, 0.97)), and inhaled steroids (32.4 percent versus 19.8 percent; OR 0.50 (0.32, 0.77)), inhaled bronchodilators (54.1 percent versus 40.1 percent; OR 0.53 (0.36, 0.78)), and supplemental oxygen (9.4 percent versus 3.0 percent; OR 0.30 (0.13, 0.73)) at time of followup. Similarly, a greater percentage of the control infants had received supplemental home oxygen at some time since NICU discharge when compared to infants who had received iNO (49.5 percent versus 38.4 percent; OR 0.65 (0.44, 0.95)). However, there was no difference between the two groups in the percent who were rehospitalized for respiratory complications, or for any reason (21.9 percent versus 22.6 percent; OR 1.03 (0.65, 1.62)) (Appendix E, Evidence Table 15).

Field, 200563 reported respiratory outcomes from assessments by pediatricians in the first year of life for 25 of 55 who had received iNO and 18 of 53 infants who had not. Three iNO infants required respiratory support day or night, and three required supplemental oxygen. Ten used bronchodilators since discharge, and five used steroids. Respiratory symptoms in the three months prior to assessment included coughing at night (8 infants) and wheezing day or night (13 infants). Nine iNO infants had respiratory signs and symptoms on exam by the pediatrician. Two control infants required respiratory support day or night, and one required supplemental oxygen. Seven had used bronchodilators since discharge and five had used steroids. Respiratory symptoms in the three months prior to assessment included coughing at night (5 infants) and wheezing day or night (11 infants). Four control infants had respiratory signs and symptoms on exam by the pediatrician (Appendix E, Evidence Table 15).

Watson, 200936 assessed the outcomes of premature infants with respiratory failure who were randomized to receive iNO versus standard therapy. One year outcomes for these infants focused on health resource utilization and neurodevelopment. Use of supplemental oxygen at home was not a primary outcome variable but was reported as: 1) percentage of infants using supplemental oxygen prior to one year corrected age; 2) percentage on oxygen at one year of age; 3) duration of supplemental home oxygen. There were no significant differences in any of these measures by study arm when the entire group was evaluated. However, when stratified into birth weight categories, the smallest infants (500 to 749 g) who did not receive iNO had an advantage; fewer required supplemental oxygen at one year corrected age (4 percent versus 11.7 percent, p-value < 0.04).

Our meta-analysis including the trials of Field, 2005 and Hibbs, 2007, showed a statistically significant lower risk for those receiving iNO therapy compared to controls in the need for bronchodilator, RR 0.75 (0.62, 0.91), and steroid therapy, RR 0.62 (0.46, 0.85), but not in wheezing, RR 1.14 (0.56, 2.32).

Respiratory health was one of the domains assessed at four to five year followup by Huddy, 2008.35 Twenty of 22 iNO infants and 15 of 16 control infants were reported to have no respiratory disability. The remaining infants in each group (2 iNO, 1 control) had mild respiratory disability (Appendix E, Evidence Table 15).

Conclusion

Our search identified twelve articles that included outcomes into early childhood. Six randomized controlled trials provided the baseline population for nine followup studies. Only Field, 2005 includes any post NICU followup among primary outcome measures, and this study included just over half of the planned sample size for this outcome. We also identified three cohort studies addressing long term outcomes. The two prospective cohort studies do not include controls for comparison. The controls in the retrospective cohort study are chosen from an earlier time period when practice standards other than just iNO use may have differed. Therefore, evidence to definitively answer any facet of this key question is not adequate.

Few individual studies and none of the meta-analyses revealed a significant association between neonatal iNO exposure and any neurodevelopmental outcome up to five years of age. For CP the two studies that did show associations conflicted in the direction of association. Tanaka, 2007 reports a decreased incidence of CP in the iNO group and Hintz reports an increase in CP in the iNO group. Both studies also had design or statistics issues that limit interpretation of the results. Mestan, 2005 reported a lower incidence of MDI < 70, NDI and the composite variable, death or NDI in those treated with iNO from Schreiber, 2003. This provides consistency, as the latter found a lower rate of CLD or death, and significant perinatal brain injury in the iNO treated infants. Of the studies that report growth parameters, Mestan, 2005 also reported the only difference in any anthropometric measure; the iNO treated infants were heavier at the time of followup. This set of results provides an incentive to pursue additional randomized controlled trials of iNO in premature infants with primary outcomes, such as neurodevelopment, that extend into early childhood.

Of the studies that reported pulmonary outcomes after NICU discharge, only Hibbs, 200844 found significant associations that favor iNO use in the NICU; iNO treated infants from Ballard, 2006 were less likely to use bronchodilators and steroids at one year of age corrected for prematurity than controls. Field, 2005 provided the only other comparable data for meta-analyses. This study increased total sample in the meta-analyses by only 10 percent and the addition of this study to the meta-analysis did not have any significant influence on the results. Meta-analyses found statistically significantly lower use of bronchodilators and steroids in the iNO treated infants at followup. Ballard, 2006 treated infants with iNO or study gas at a later chronological age than most RCTs (at 7 to 21 days) and for the longest duration, a minimum of 24 days. This is compelling evidence, but it is not sufficient to recommend routine use of iNO for protection against chronic respiratory illnesses of childhood. It does, however, warrant directing focus to additional RCTs of iNO use in premature infants in the NICU and considering that the timing of initiation and duration of therapy may play an important role in outcome. Design of future studies should focus on early childhood outcomes, with definitive and objective outcome measures.

Key Question 4. Does the effect of iNO therapy on BPD and/or death or neurodevelopmental impairment vary across subpopulations of premature infants?

Major Findings

  • There is insufficient evidence to determine whether the effect of iNO therapy on mortality, BPD, or motor impairment differs by the birth weight of the treated infants.
  • There is insufficient evidence to evaluate the relationship between iNO therapy and infant sex, race/ethnic group, gestational age, or socioeconomic status.
  • There are no published data available to evaluate the association between iNO therapy and, antenatal steroids, chorioamnionitis, multiple birth, or growth restriction.
  • There is insufficient evidence concerning the relationship between iNO therapy and the severity of illness.
  • There is insufficient evidence that iNO therapy improves outcome of infants suffering respiratory failure from pulmonary hypoplasia, respiratory distress syndrome or pulmonary hypertension.
  • There is no consistent pattern of infants that respond to iNO therapy and those that do not.

Detailed Analysis

Six randomized controlled trials,34, 37, 40, 58, 63, 62 four with long term followup,30, 36, 56, 57 and seven other studies38, 68, 69, 70, 73, 74, 77 addressed one or more subpopulations of interest in this Key Question (Table 9).

Table 9. Summary of outcomes for RCTs addressing KQ4.

Table 9

Summary of outcomes for RCTs addressing KQ4.

Four RCTs investigated whether iNO therapy has a differential effect by birth weight37, 40, 58, 34; three of the trials34, 37, 40 reported long term followup.30, 36, 57 Birth weight subgroup analyses were planned a priori in two trials34, 37, 58 and were done post hoc for the other two trials.40, 58 Three trials enrolled infants at ≤ three days of age,40, 58, 92 while the fourth RCT enrolled infants at seven to 21 days of age.34

Only small numbers of trials have addressed the effect of iNO therapy on other subpopulations including the severity of infant illness, as measured by the oxygenation index (OI)40, 58, 63 or respiratory severity score,34 race,34, 37, 57, 62 sex,34, 57, 62 gestational age,62 pulmonary hypertension,38 and pulmonary hypoplasia.69, 77

Descriptions of studies that evaluate iNO therapy in subgroups of infants by demographic characteristics are reviewed first below. Studies that evaluated iNO therapy by severity of illness indicators and causes for respiratory failure follow (Appendix E, Evidence Tables 3 and 4; Table 9).

Birth weight. The evidence for the effect of iNO is presented in standard birth weight groupings: < 750 g, 750 to 999 g, 1000 to 1250 g, ≤ 1000 g, > 1000 g, and others. For individual studies, the birth weight stratum may vary slightly from the category heading. For instance, results for a study using the stratum ≤ 750 g are included under the heading < 750 g. Ballard, 200634 is an exception as infants were categorized into birth weight groups of 500 to 799 g and 800 to 1250 g. This trial has been reviewed in the birth weight category of < 750 grams and 1000 to 1250 grams birth weight

Birth weight < 750 g. In two trials,30,71 including one with 384 infants with birth weight between 500 g and 749 g, there was no difference in mortality in the NICU,37 or survival without chronic lung disease58 between those treated with iNO and controls. In followup to a third trial,40 mortality at 18 to 22 months was significantly higher in the iNO group compared with the controls (73 percent versus 56 percent; p-value = 0.01).30 A fourth trial, Ballard 200634, reported no significant difference between infants treated with iNO and controls in survival without CLD at 36 wks PMA (RR 1.26 (0.98, 1.62)) or death (RR 1.02 (0.96, 1.08)), among infants with birth weight of 500 to 799 g80 (Appendix E, Evidence Table 16).

No difference was reported in the incidence of BPD at 36 weeks PMA between groups for the three trials that reported the outcome.34, 37, 40 At one year corrected age, Watson 200936 reported more infants treated with iNO in Kinsella, 200637 remained on oxygen compared with control infants (11.7 percent versus 4 percent, p-value = 0.04). (Appendix E, Evidence Table 16).

No meta-analyses were conducted for this Key Question because of the differences in the definitions of subgroups across studies and the reported outcomes measured.

Similar rates of the composite outcome death or BPD at 36 weeks PMA were reported for iNO treated infants and controls in all three studies that reported this outcome.34, 37, 58 In one followup study, the combined outcome of death or an oxygen requirement to one year of age occurred in 37 percent of infants in each group36 (Appendix E, Evidence Table 16).

Inhaled nitric oxide therapy did not improve neurodevelopmental outcome in this birth weight category. Neurodevelopmental impairment (NDI) was similar between iNO treated infants and controls when measured by Hintz 2007 at 18 to 22 months corrected age (NDI defined as including any of the following: moderate to severe CP, blind, deaf, MDI < 70, or PDI < 70),36 by Walsh, 2010 at 24 months corrected age (NDI defined as moderate or severe CP, bilateral blindness, bilateral hearing loss requiring amplification, or score <70 on the Bayley Scales MDI or PDI) (RR 0.85 (0.67, 1.08))57, and by Watson 2009 at one year of age corrected for gestational age at birth (NDI defined as including any of CP, blindness, severe hearing loss, MDI < 70 or PDI < 70).30 In two followup studies30, 36 the composite outcome death or NDI was similar between the groups. The composite death or moderate to severe CP occurred more frequently in the iNO treated infants than controls (81 percent versus 62 percent, p-value = 0.0039), in one study.30 (Appendix E, Evidence Table 16).

Birth weight 750 to 999 g. No differences were reported between iNO and control infants in this subgroup with respect to mortality, BPD, the combined outcome of death or BPD, neurodevelopmental impairment (NDI),37, 40 or survival without BPD58 in the three studies. Watson, 2009 reported that iNO treated infants had lower rates of death or NDI at one year compared with controls (32.1 percent versus 44.4 percent, p-value=0.04), as well as a decreased rate of the combined outcome of death, on oxygen, or NDI at one year corrected age (32.9 percent versus 45.1 percent, p-value = 0.04)36 (Appendix E, Evidence Table 16).

Birth weight <1000 g. Only one study examined this birth weight subgroup, using post hoc analyses. The iNO treated infants had a higher mortality rate than the control group (62 percent versus 48 percent, RR 1.28 (1.06, 1.54)), but they also had a higher rate of severe (Grade 3 or 4) IVH (43 percent versus 33 percent, RR 1.40 (1.03, 1.88)). No difference was found in the incidence of BPD, or the composite outcome death or BPD40 (Appendix E, Evidence Table 16).

At followup to 18 to 22 months corrected age, the iNO group had a higher rate of death (98/152, 64 percent) than the control group (79/152, 52 percent; p-value=0.04). Those treated with iNO also had a 22 percent greater rate of death or moderate to severe CP at 74 percent (111/151) compared to 59 percent (89/152) in the control group (RR 1.22 (1.05, 1.43) p-value = 0.01) 30 (Appendix E, Evidence Table 16).

Birth weight 1000 to 1250 g. The largest RCT that described birth weight subgroups and outcomes is Kinsella, 2006.37 For this higher birth weight stratum, Kinsella reported a significant reduction in the combined outcome of death or BPD for the iNO treated infants (38.5 percent versus 64.1 percent, RR 0.60 (0.42, 0.86)), as well as a lower rate of BPD alone (29.8 percent versus 59.6 percent, RR 0.50 (0.32, 0.79)), although there was no difference in death alone. In followup at one year corrected age, there were no significant differences in the incidence of NDI, death, subjects on oxygen, or any composite outcomes36 (Appendix E, Evidence Table 16).

In the Ballard, 2006 RCT,34 which stratified infants across the wider birth weight category of 800 to 1250 g, there was no significant difference for iNO treated infants compared with controls for death (RR 1.00 (0.95, 1.06)) or survival without BPD at 36 wks PMA (RR 1.25 (0.88, 1.79)).80 There was also no significant difference between iNO treated infants and controls for BPD alone (51.5 percent versus 61.5percent) nor death or survival with BPD (54.6 percent versus 64.8 percent).80 In followup at two years corrected age,57 there was no difference between groups in the incidence of NDI (RR 1.07 (0.76, 1.50)).

Other birth weight groups including infants larger than 1250 g. No differences were reported for any outcome in studies that reported birth weights of 1000 to 1500 g40,58 or >1500 g.58 However, in post hoc analyses for the subgroup of infants with birth weight > 1000g, Van Meurs, 200540 found a lower rate of the composite outcome of death or BPD for the iNO treated group compared to controls (50 percent versus 69 percent, p-value = 0.03; RR 0.72 (0.54, 0.96)), but no difference in death or BPD alone (Appendix E, Evidence Table 16).

We determined that meta-analyses of trials reporting outcomes by birth weight subgroups would not be performed due to the differences in definitions of birth weight categories, the marked variability in iNO administration and differences in outcomes reported in these few trials.

Gestational age. Mercier, 201062 assessed the relationship between iNO and gestational age at birth in infants born at less than 29 weeks. A similar incidence in survival without BPD at 36 weeks PMA was reported for infants treated with iNO therapy and controls resulting in a risk ratio for those with gestational age <26 weeks of RR 1.14 (0.71, 1.82), and for those with gestational age ≥26 weeks of RR 0.94 (0.64, 1.38) (Appendix E, Evidence Table 16).

Sex. Two RCTs commented on the association of iNO therapy and an infant’s sex. In post hoc analysis, Ballard, 200634 stated that there was no difference in the response to iNO according to sex, but no data were shown. In the two year followup to Ballard, 2006, Walsh, 2010 reported there was no interaction between treatment with iNO and infant sex for the composite outcome NDI.57

Mercier, 201062 also showed no treatment effect by sex with similar relative risks of survival without BPD at 36 weeks PMA among girls treated with iNO compared with controls, RR 1.18 (0.76, 1.83), and boys treated with iNO compared with controls, RR 0.85 (0.58, 1.26) (Appendix E, Evidence Table 16).

Race/ethnicity. Kinsella, 200637 performed post hoc analyses and found no significant effect of race or ethnic group on the composite outcome of death, ICH or PVL following iNO treatment. In a post hoc analysis by Ballard, 200634 the effect of iNO did not differ significantly according to race or ethnicity (p-value = 0.06).75 The risk ratios for survival without BPD at 36 weeks PMA by individual race follow: whites RR 1.06 (0.76, 1.47), blacks RR 1.72 (1.20, 2.47), Hispanics RR 1.66 (1.06, 2.59), and other 0.54 (0.25, 1.14).80 Walsh 2010, in two year followup to the Ballard, 2006 trial, found no significant interaction between iNO treatment and race, white infants versus non-white infants, in NDI.57 Mercier 201062 reported no difference between those receiving iNO and controls in survival without BPD for black infants, RR 1.49 (0.61, 3.65), or non-black infants, RR 0.94 (0.69, 1.28). (Appendix E, Evidence Table 16). None of the studies reporting on race/ethnicity were powered to address these subgroups.

Socioeconomic status. In the only study to consider socioeconomic indicators of outcome, Walsh, 2010 reported no statistically different risk of NDI at two year followup for infants treated with iNO and controls when mothers had less than a high school education, RR 0.77 (0.46, 1.30), compared to those that had a high school education or greater, RR 0.92 (0.72, 1.17)57; no data was provided in the original trial34 (Appendix E, Evidence Table 16).

Other subgroups. There were no trials that specified outcomes by subgroups of exposure to antenatal steroids, chorioamnionitis, multiple births, and small for gestational age.

Description of trials based on severity of illness.

Oxygenation index. Three RCTs used the oxygenation index (OI, OI = mean airway pressure in cm H2O x fraction of inspired O2 x 100)/postductal arterial partial pressure of O2 (PaO2) in mm Hg) as a surrogate measure of severity of illness.

The study of Van Meurs, 200540 required an OI ≥ 10 on two consecutive arterial blood gases (ABGs) for study entry. Following the first interim analysis, due to a higher than expected mortality rate in both treatment and control arms, the respiratory criteria for study entry were revised to an OI ≥ five followed by an OI ≥ 7.5. In infants with birth weight 401 to 1500 g, the mean OI (SD) at randomization was 23±17 for the iNO treated infants and 22 ± 17 for the controls. Post hoc analysis indicated no interaction between iNO treatment and OI stratum. The risk of death, BPD, and death or BPD were similar between the iNO treatment and control groups for those with a median OI ≤ 17 and for those with OI > 17 at the time of randomization. Severe IVH or PVL rates were similar between groups within the OI strata40 (Appendix E, Evidence Table 16).

Field, 2005 reported on a total cohort of 108 subjects with a notably high severity of illness as assessed by OI. At study entry, the iNO treated group had a median (IQR) OI of 32.9 (22.2, 49.8), and 55 percent had an OI > 30 as compared to the control group median OI 31.9 (17.4, 51.8), and 53 percent with an OI > 30. When primary outcomes were stratified by OI < 30 or ≥ 30, there were no significant differences in death or severe disability (defined as no/minimal head control or inability to sit unsupported or no/minimal responses to visual stimuli) (RR 0.99 (0.76, 1.28), p-value = 0.62); death or supplemental O2 at expected date of delivery (RR 0.83 (0.68, 1.02), p-value = 0.87); or death or supplemental O2 at 36 weeks PMA (RR 0.98 (0.87, 1.12), p-value = 0.81) in the iNO group compared to controls63 (Appendix E, Evidence Table 16).

In Schreiber, 2003,58 a post hoc analysis was performed, stratified by OI as a measure of severity of illness. For the group of iNO treated infants with OI < 6.94 (median), there was a significantly decreased risk of the composite outcome death or survival with CLD compared with the placebo group (36 percent versus 67.4 percent; RR 0.53 (0.35, 0.81)). There was no significant difference for the subgroup with OI ≥ 6.94. Mestan, 200956 reported the neurodevelopmental outcomes at a corrected age of two years for the cohort of survivors (N = 138). In a post hoc analysis, in comparison with the placebo group the iNO treated group with initial OI < 6.94 had no significant difference in abnormal neurodevelopmental outcome (defined as either disability (CP, bilateral blindness, or bilateral hearing loss) or delay (MDI < 70 or PDI <70), RR 0.52 (0.26, 1.01), but for the iNO treated infants with initial OI ≥ 6.94 there was 62 percent lower risk for abnormal neurodevelopmental outcome, RR 0.38 (0.16, 0.93) (Appendix E, Evidence Table 16).

We opted not to undertake a meta-analysis since these three RCTs had such a wide disparity in the OI criteria used, rendering them much less clinically comparable. The relatively low median OI in the Schreiber trial58 reflects a population of infants presumably less critically ill than those in the Van Meurs trial40 with a median OI of 17, and markedly less critically ill then those infants in the Field trial63 with more than half having an OI > 30. The Mestan study is the only one to report neurodevelopmental outcomes by OI.

Respiratory severity score. In Ballard, 2006,34 a simplified respiratory severity score was used, calculated as the mean airway pressure x FiO2, since actual PaO2 (needed to calculate the OI) was often not available. At study entry, the median severity score was 3.5 for both iNO treated and control groups, and was noted to be equivalent to an OI in the range of five to nine. In post hoc analyses, there was no interaction between the severity score at study entry and treatment for the outcome survival without CLD at 36 weeks PMA.34, 80 Followup at two years corrected age showed no difference in NDI between iNO treated and control infants if the respiratory severity score was < 3.5, RR 0.93 (0.69, 1.26) or ≥ 3.5, RR 0.93 (0.72, 1.19)57 (Appendix E, Evidence Table 16).

Other measures of severity of illness. We found no trials that studied the effect of iNO therapy by subgroups defined by oxygen requirement alone.

Description of trials based on causes of respiratory failure.

Respiratory distress syndrome. In Field, 2005, the primary outcome measures were stratified by principal diagnoses, defined as acute preterm lung disease (presenting immediately after birth and randomized at ≤ 3 days of age), chronic preterm lung disease (presenting immediately after birth and randomized for continuing problems after 3 days of age), and other (developed lung disease after recovering from an initial respiratory problem). There were no differences between iNO treated infants and controls in death or supplemental O2 at 36 weeks PMA, RR 0.98 (0.87, 1.11); death or supplemental O2 at the expected date of delivery, RR 0.83 (0.68, 1.01); or death or severe disability, RR 0.99 (0.76, 1.28)63 (Appendix E, Evidence Table 16).

Pulmonary hypoplasia. In very low birth weight preterm infants, pulmonary hypoplasia can occur following maternal preterm premature rupture of membranes (PPROM) > five days with subsequent oligohydramnios, and may further be complicated by persistent pulmonary hypertension. Two groups of investigators conducted retrospective analyses of infants with suspected pulmonary hypoplasia. In a subset analysis of infants with suspected pulmonary hypoplasia from the two Van Meurs trials, Chock, 200977 compared six infants exposed to iNO with six controls; the infants were similar at baseline. There was no statistically significant difference between iNO treated infants and controls in death (33 percent versus 67 percent, p-value = 0.57), BPD at 36 weeks PMA in the seven survivors (2/5 (40 percent) versus 2/2 (100 percent), p-value = 0.43)), or death or BPD (50 percent versus 100 percent, RR 0.50 (0.22, 1.11)). At 18 to 22 months followup, none of the four surviving iNO treated infants assessed had NDI (defined as moderate to severe CP, blindness, or deafness); the two survivors from the placebo group were lost to followup, and thus no comparisons were made. Uga, 200469 compared eight infants treated with iNO to 10 controls. All eight infants treated with iNO survived to 28 days compared to 5/10 control infants (p-value < 0.05). The groups had similar rates of BPD (undefined) (Appendix E, Evidence Table 16).

A meta-analysis was not performed for these two retrospective cohort studies with very limited numbers of infants enrolled, and widely different time points for death, i.e., prior to discharge home or within 365 days in hospitalized infants77 versus seven and 28 days.69

Pulmonary hypertension. In persistent pulmonary hypertension of the newborn (PPHN), the pulmonary vascular resistance remains elevated in the newborn period, and it is the primary U.S. FDA approved indication for iNO in the term and near term infant population. In a retrospective case control study of 31 singleton preterm infants at median 25 weeks gestational age (IQR 24 – 28 weeks) with clinical pulmonary hypertension confirmed by echocardiography, Tanaka, 2007 reported that at three years of age 2/9 (22.2 percent) infants with CP had been treated with iNO compared to 14/22 (63.6 percent) infants without CP38 (Appendix E, Evidence Table 16).

iNO responders compared to nonresponders. Four studies reported primary outcomes by the presence or absence of response to iNO therapy. Yadav, 1999 reported results from a retrospective study of iNO given to 41 preterm infants with a mean OI of 40 on maximal medical therapies. Response to an initial 10 ppm iNO was defined as a decrease in OI by ≥ 10 at one hour of treatment. The 26 responders and 15 nonresponders were similar with respect to birth weight, gestational age, and OI at the start of treatment. Death was reported as 11/26 (42 percent) for responders but 14/15 (93 percent) for nonresponders. In a multivariable model, early response to iNO was associated with survival to discharge (p-value = 0.01)73 (Appendix E, Evidence Table 16).

Banks, 1999 studied iNO usage in severe BPD in 16 ventilator dependent preterm infants more than a month old (range 1 to 7 months). In this open label non controlled trial, iNO was administered at 20 ppm for the first 72 hours, then titrated slowly (median duration 27 days) for responders or discontinued in 24 hrs for nonresponders. Non response was defined as a 10 percent increase in oxygen requirement, a PaCO2 of 70 mm Hg on baseline ventilator settings, worsening chest x ray, or a methemoglobin of > five percent within 72 hours of starting iNO therapy. Mortality for the overall cohort was 44 percent (7/16). For iNO responders, 4/11 (36 percent) died over the range of 11 days to five months, while in the nonresponder group 3/5 (60 percent) died and the two survivors remained ventilator dependent70 (Appendix E, Evidence Table 16).

Kumar, 2007 performed a retrospective chart review of preterm infants < 37 weeks gestational age at birth with pulmonary hypertension treated with iNO. Pulmonary hypertension was diagnosed by echocardiography within the first four weeks of life; iNO treatment at doses of 5 to 15 ppm was at the discretion of the clinical team after an infant failed standard medical management. Within the gestational age range of interest in this evidence report, ≤ 34 weeks gestation, response to iNO was least likely to occur in the most immature infants: only 1/6 (16 percent) of infants < 29 weeks responded to iNO, defined as an increase in postductal PaO2 of 20 mm Hg or greater within 30 minutes without any change in inspired oxygen, while 5/6 (83 percent) of infants 29 to 31 weeks gestation, and 5/6 (83 percent) of infants 32 to 34 weeks gestation responded to iNO68 (Appendix E, Evidence Table 16). Mortality was significantly higher for the non responders (6/8, 75 percent) compared to iNO responders (4/15, 26 percent) (p < 0.04) (Appendix E, Evidence Table 16).

In a pilot study of the European iNO Registry, Dewhurst 2010,74 reported the outcome for 44 preterm infants <34 weeks gestational age at birth. Infants with congenital heart disease were excluded. Infants were treated with a median starting dose of iNO of 20 ppm (range 3.3 to 25 ppm), and a median maintenance dose of 10 ppm (range 0.7 to 25 ppm). Response to iNO was defined as a 15 percent reduction in the baseline OI within 30 to 60 minutes of starting therapy; 26 infants responded and eight did not. Infants that responded to iNO were younger than non responders (median (IQR) gestational age at birth, 26 (25 to 29) weeks versus 29 (27 to 30) weeks, p-value = 0.043) and had a significantly higher baseline oxygen requirement (median (IQR) Fi02 1.0 (0.9, 1.0) responders versus 0.8 (0.5, 1.0) non responders; p-value = 0.021). Birth weight, age at starting iNO, starting dose, and baseline OI were similar between responders and non responders. There was no difference in mortality: 12 of the 21 responders for whom complete data were available died; 5/8 non responders died (Appendix E, Evidence Table 16).

No meta-analysis for the iNO responder versus non responder studies was undertaken due to the wide variations in iNO dosage and timing, as well as the variability and incomplete description of the diagnoses underlying the respiratory failure in these preterm cohorts.

Conclusions

For the question of whether iNO therapy has an effect on the major outcomes of interest (death and/or BPD or neurodevelopmental impairment) across various subpopulations of premature infants, we reviewed 17 studies which included six original RCTs,34, 37, 40, 58, 63, 74 with four followup studies,30, 36, 56, 57, 77 and seven other studies,38, 68, 69, 73, 70, 74, 77 As noted in the conclusion of Key Question 1, some of the studies that reported no significant differences in rates of death, BPD at 36 weeks PMA, or the composite outcome death or BPD at 36 weeks PMA for the overall cohort did find statistically significant subgroup differences. For example, two studies37, 40 report decreased rates of one or more of the major outcomes among infants with birth weight > 1000 grams.36, 56 The lack of consistency in defining or subdividing certain subgroups, (e.g., by birth weight, or oxygenation index) hampered the ability to answer this Key Question. In addition, many of the subgroup analyses performed were by post hoc analyses (e.g., birth weight, race), increasing the need for cautious interpretation of results. Some of the specific subpopulations of interest had little or no outcomes related data at all. Based on the current body of evidence, no definitive and generalizable conclusions may be made about iNO treatment in specific subpopulations. Future research is needed to examine the role of iNO treatment in these and alternative subgroups, so as to more clearly define populations of preterm infants which may benefit most from this therapy.

Key Question 5. Does the effect of iNO therapy on BPD and/or death or neurodevelopmental impairment vary by timing of initiation, mode of delivery, dose and duration, or concurrent therapies?

Major Findings

  • There is insufficient evidence to determine if initiating iNO therapy for acute respiratory distress at ≤ three days reduces the risk of death or bronchopulmonary dysplasia (BPD) at 36 weeks PMA, or death and neurodevelopmental disability at one year of age, corrected for gestational age at birth.
  • In infants with developing BPD, there is insufficient evidence to determine if treatment with iNO during the second week after birth improves survival without BPD compared with treatment during the third week after birth.
  • There is insufficient evidence to determine the effect of delivery of iNO by high frequency ventilation on either death or BPD, or neurodevelopmental outcome compared with conventional ventilation.
  • There is insufficient evidence to support an optimal dose of iNO or duration of exposure to improve outcome or prevent harm.
  • There is insufficient evidence to determine the effect of iNO with concurrent therapy.

Detailed Analysis

Fourteen RCTs reported in 21 papers addressed this key question. Two trials investigated the timing of the initiation of iNO therapy,34, 63 and two the mode of drug delivery, conventional or high frequency ventilation.40, 58 The dose of iNO varied considerably among the 14 studies. To examine the effect of dose on the primary outcomes, studies were categorized into those that administered iNO at only 5 ppm,37, 59, 62 those that delivered a maximum dose of 10 ppm,39, 40, 58, 61, 67 and those that gave 20 ppm or titrated the dose to the patients’ response.34, 60, 63-66 Duration of iNO exposure also varied considerably among the 14 studies, from three to four days64 to a minimum of 24 days.34 As the majority of the studies administered iNO therapy until extubation, the evidence for the effect of duration of exposure of iNO on the primary outcomes of BPD and/or death or neurodevelopmental impairment could not be evaluated. Only two studies explicitly considered concurrent therapies, specifically systemic steroids, on the effect of iNO treatment.64, 76

All trials reported death or survival, although the time of ascertainment of the outcomes varied across studies. If death or survival was reported at multiple endpoints in either the original study or in a long term followup publication (e.g., before NICU discharge, and at one year of age, corrected for gestational age at birth), the data are included in this evidence report. Seven of the randomized trials have reported long term followup.30, 35, 36, 56, 57, 76, 78 The followup studies have varying definitions of neurodevelopmental impairment (NDI), and use different measures of developmental progress, making direct comparison difficult. For questions concerning some subgroups (timing of initiation of iNO therapy, and mode of iNO delivery) the analyses are post hoc, therefore the results must be considered exploratory, as infants were neither randomly assigned to the subgroup, nor was the study powered to consider the variable (Appendix E, Evidence Tables 3 and 4; Table 10). The differences in treatment protocols in studies reporting on subgroups of infants may make pooled estimates of the effect of iNO therapy spurious, so meta-analyses were performed only with RCTs using similar dosing regimens.

Table 10. Summary of outcomes for RCTs addressing KQ5.

Table 10

Summary of outcomes for RCTs addressing KQ5.

Timing: Early versus late iNO administration. Two populations of preterm infants have been treated with iNO, those with early acute respiratory distress (immediately after birth), and those with evolving BPD. Early treatment generally begins within the first three days after birth in an effort to improve oxygenation in infants with acute hypoxemia, as in respiratory distress syndrome (RDS) or pneumonia. Late treatment may begin any time after three days, with evidence of progressive respiratory failure. Theoretically, late treatment avoids exposing an infant to iNO who would otherwise have resolving RDS during the first week. The goal of both strategies is to prevent the development of BPD and all its sequelae. Although the age at initiation of iNO therapy was available for all trials, aggregating studies into meaningful categories was problematic; RCTs started iNO therapy at < 48 hours, < 72 hours, < 96 hrs, four to 120 hours, < seven days, seven to 21 days, and < 28 days of age. Some trials also had varying additional entry criteria concerning severity of illness, further complicating the ability to combine studies into clinically meaningful groups. Because of this variability we did not conduct meta-analyses; instead we report the results of two RCTs that specifically considered the timing of the initiation of iNO therapy, both in post hoc analyses (Appendix E, Evidence Table 17).

In a small sample, Field, 200563 found a similar risk of death or BPD at 36 weeks PMA in infants treated with iNO within three days of birth (25/38, 66 percent) and those treated at four to 28 days (12/17, 71 percent; RR 0.98 (0.87, 1.11)). There was an advantage to early iNO administration when death or BPD was measured at the expected date of delivery (61 percent versus 94 percent when iNO was initiated at 4 to 28 days), but the difference was attenuated after adjustment for diagnosis (acute lung disease beginning at birth and treated at ≤ three days, chronic lung disease with respiratory distress at birth and continuing at four to 24 days; and other respiratory distress after recovery from an initial respiratory problem), and severity of illness (OI ≤ 30 versus > 30), RR 0.83 (0.69, 1.01). The time of initiation of iNO had no effect on death or severe disability, defined as no/minimal head control or the inability to sit unsupported or no/minimal response to visual stimuli, at one year of age corrected for gestational age at birth (RR 0.99 (0.76, 1.28)). No data were provided on the median time that iNO therapy was initiated in the early or late group, but the median (IQR) age of initiation for all infants receiving iNO therapy in the study was 1 (0, 6) days, making it unlikely that the groups were very different (Appendix E, Evidence Table 17).

In a large sample of nearly 600 infants with developing BPD, Ballard, 200634 reported a similar incidence of death at 36 weeks PMA between the iNO and placebo groups for those entering treatment at seven to 14 days (10.7 percent iNO versus 11.3 percent placebo), or those entering treatment at 15 to 21 days (6.6 percent iNO versus 5.8 percent placebo) 34 However, the likelihood of survival without BPD increased for infants beginning treatment at 7 to 14 days (RR 1.91 (1.31, 2.78)), a result that was not observed in those beginning treatment later, at 15 to 21 days (RR 0.99 (0.77, 1.28)).34 This result may have been observed because of damage already done to the developing lung before late enrollment (Appendix E, Evidence Table 17).

In both of these studies analyses were conducted post hoc, so neither study was powered to find a statistically significant difference between the groups (Appendix E, Evidence Table 17).

Mode of drug delivery. Two randomized controlled trials, Van Meurs, 200540 and Schreiber, 200358 reported outcomes for infants treated with iNO and conventional mechanical ventilation compared with those treated with iNO and high frequency ventilation. Patients were randomly assigned to ventilation strategy in one trial,58 but in the other, analyses were done post hoc.40 Both studies reported neurodevelopmental followup to 18 to 24 months of age, corrected for gestational age at birth.30, 56 Patients in the two trials differed by birth weight inclusion criteria (401 to 1500 grams,40 ≤ 2000 grams58) (Appendix E, Evidence Table 17).

Schreiber, 200358 found no difference in the combined outcome of death or BPD at 36 weeks PMA between infants randomized to treatment with iNO and conventional ventilation (RR 0.61 (0.41, 0.90)) compared with placebo and those randomized to iNO and high frequency ventilation (RR 0.92 (0.67, 1.26)) compared to placebo. Among survivors, the risk of abnormal neurodevelopmental outcome, defined as disability (CP, bilateral blindness, or bilateral hearing loss) or developmental delay (a score of <70 on the Bayley Scales of Infant Development II, but no disability) was not statistically different between high frequency ventilation and conventional ventilation, RR 0.92 (0.58, 1.46)56 (Appendix E, Evidence Table 17).

In post hoc analysis, Van Meurs, 200540 reported that iNO delivered by conventional mechanical ventilation was associated with a 46 percent increase in the risk of death before discharge to home or within 365 days of birth among infants still hospitalized compared with placebo (RR 1.46 (1.10, 1.92)). The risk remained elevated at 18 to 22 months of age (RR 1.37 (1.05, 1.79)).30 There was no increase in death, at either time, among infants treated with iNO delivered by high frequency ventilation compared to placebo. The risk of developing BPD at 36 weeks PMA was similar if iNO was delivered by conventional ventilation (RR 0.90 (0.65, 1.24)), or high frequency ventilation (RR 0.89 (0.72, 1.10)) (Appendix E, Evidence Table 17). Motor development was impaired at 18 to 22 months of age, in those treated with iNO and conventional ventilation (CP RR 1.29 (1.03, 1.60)), and there was an increased risk of death or moderate to severe CP (moderate CP was defined as the ability to sit independently or with support but cannot independently ambulate; severe CP was defined as unable to sit or walk even with support), RR 1.29 (1.03, 1.60)30 (Appendix E, Evidence Table 17). The risk of disability, defined as moderate/severe CP, bilateral blindness, deafness, or MDI or PDI < 70, was not affected by mode of iNO delivery, nor was the combined outcome of death or disability (high frequency ventilation RR 0.97 (0.70, 1.35); conventional ventilation RR 1.07 (0.64, 1.80)30 (Appendix E, Evidence Table 17).

Although treatment with iNO and conventional ventilation was associated with some measures of adverse outcome, the estimates of harm are the result of post hoc analyses and so must be considered cautiously (Appendix E, Evidence Table 17). We did not perform a meta‐analysis for mode of delivery as only one study prospectively randomized infants to conventional versus high frequency ventilation and the other generated comparisons by post hoc analyses.

No data are available on other methods of iNO delivery, such as high or low flow nasal cannula, or continuous positive airway pressure.

Dose of iNO. The initial dose of iNO varied from 5 ppm to 20 ppm among the 14 randomized trials. Data in preterm animal models of RDS indicates improvement in oxygenation in this range.93 Fear of adverse side effects, specifically bleeding with resulting IVH, resulted in limitation of iNO exposure in some studies (Appendix E, Evidence Table 17). For this review, studies were grouped as follows: dose restricted to 5 ppm; dose restricted to 10 ppm; dose titrated to response with a maximum of dose 20 ppm to 40 ppm, or dose given as 20 ppm. The effect of iNO doses on the primary outcomes are reviewed below. Meta-analyses were conducted with trials that reported death in the NICU at 36 weeks PMA or later; trials that reported death at seven days or 28 days were excluded. Separate meta-analyses were done that excluded Ballard 2006, because it was the only RCT that delivered iNO for a prolonged period of time (minimum 24 days); the results did not differ from what is reported.

Death. When the dose of iNO was restricted to 5 ppm there was no difference in death at 36 weeks PMA in two trials with 800 infants enrolled in each, RR 0.79 (0.61, 1.03) 37 and RR 1.34 (0.92, 1.95)62, or in death prior to hospital discharge, RR 0.90 (0.58, 1.40), in one trial with 80 infants59 (Appendix E, Evidence Table 17). The pooled estimate of the risk of death showed no significant differences in treatment with this dose of iNO, RR 0.97 (0.70, 1.35) (Figure 16).

This figure is a forest plot of 11 RCTs used to calculate whether pooled data on the impact of iNO dose (stratified at 5, 10 and 20 ppm) on death in the NICU, in preterm infants, is significant.

Figure 16

Meta-analysis for dose-stratified death, including only studies that reported death in the NICU at 36 weeks PMA or later.

The risk of death was similar in infants that received iNO therapy at a maximum dose of 10 ppm compared with those receiving placebo in five randomized controlled trials39, 40, 58, 61, 67 (Appendix E, Evidence Table 17). In the two studies by Van Meurs the risk of death before discharge home or within 365 days for those still hospitalized was not improved with iNO therapy in infants weighing 401 to 1500 grams birth weight, RR 1.16 (0.96, 1.39),40 or those with birth weight greater than 1500 grams, RR 1.34 (0.45, 4.00) 39 (Appendix E, Evidence Table 17). Three other studies found no significant decrease in death in the NICU (RR 0.68 (0.38, 1.20))58; 20 percent iNO versus 30 percent placebo, p-value = 0.49467, or at 28 days (41 percent iNO versus 31 percent placebo, no significant difference)61 (Appendix E, Evidence Table 17). Meta‐analysis of the four studies that reported death in the NICU at 36 weeks PMA or later confirmed no statistically significant difference in death for infants treated with 10 ppm iNO compared to controls, RR 1.00 (0.73, 1.38) (Figure 16).

No study found a difference in death between infants that were treated with iNO at 20 ppm or iNO titrated to response and those given standard care, regardless of whether the outcome was measured at seven days, 66 during NICU hospitalization,60, 64, 65 at 36 weeks PMA,34 40 weeks PMA,34 44 weeks PMA, 34 at one year of age corrected for gestational age at birth63, or four to five years of age35 (Appendix E, Evidence Table 17). The pooled estimate of the relative risk of death at 36 weeks PMA or later during NICU hospitalization showed no statistically significant difference between infants given iNO therapy delivered at 20 ppm or titrated to response and controls, RR 0.91 (0.63, 1.30) (Figure 16).

The RR is similar across each dose category, with little heterogeneity, suggesting that the effect of iNO on the outcome of death does not vary by dose.

BPD at 36 weeks PMA. Bronchopulmonary dysplasia developed as frequently in those treated with 5ppm iNO as those treated with standard therapy when measured at 36 weeks PMA37, 59, 62 and when measured prior to hospital discharge59 (Appendix E, Evidence Table 17) The pooled estimate of the risk of BPD for the three trials that reported the outcome at 36 weeks PMA was RR 0.94 (0.87, 1.02) (Figure 17).

This figure is a forest plot of 11 RCTs used to calculate whether pooled data on the impact of iNO dose (stratified at 5, 10 and 20 ppm) on BPD at 36 weeks PMA, in preterm infants, is significant.

Figure 17

Meta-analysis for dose-stratified BPD at 36 weeks PMA.

The risk of BPD at 36 weeks PMA was mixed when iNO was given at a maximum dose of 10 ppm. The largest trial, conducted by Van Meurs 2005, with more than 420 infants, reported a 24 percent decrease in the risk of BPD compared to controls, RR 0.76 (0.58, 0.98).40 Three other trials, with a combined enrollment of 276 infants, reported no significant difference between infants treated with iNO and controls39, 40, 58, 67 (Appendix E, Evidence Table 17). A meta‐analysis with all four RCTs found a 25 percent reduction in the risk of BPD at 36 weeks for infants treated with iNO at 10 ppm compared to controls, RR 0.75 (0.61, 0.91) (Figure 17).

BPD at 36 weeks PMA was common, affecting one third to one half of all infants; the rate was not different between infants treated with 20 ppm or iNO titrated to response and those receiving standard care in the four RCT that used this dosing strategy.34, 63, 64, 65 The rate of BPD was also similar between the groups when defined as an oxygen requirement measured at 40 weeks PMA (22 percent iNO versus 29 percent placebo)34, 44 weeks PMA (9 percent iNO versus 12 percent placebo),34 and at one year, corrected for gestational age at birth (15 percent iNO versus 6 percent placebo)60 (Appendix E, Evidence Table 17). A meta-analysis confirmed no significant difference in the risk of BPD at 36 weeks with iNO therapy at this dose, RR 1.00 (0.74, 1.34) (Figure 17).

Death or BPD. There was no reduction in the composite outcome of death or BPD with 5ppm iNO compared with controls, when measured at 36 weeks PMA,37, 62 or prior to hospital discharge 59 (Appendix E, Evidence Table 17). Meta-analysis with these three trials resulted in a RR 0.94 (0.88, 1.01) (Figure 18).

This figure is a forest plot of 11 RCTs used calculate whether pooled data on the impact of iNO dose (stratified at 5, 10 and 20 ppm) on death or BPD at 36 weeks PMA, in preterm infants, is significant.

Figure 18

Meta-analysis for dose-stratified death or BPD.

Evidence concerning the risk of death or BPD was mixed among the trials that gave 10 ppm iNO (Appendix E, Evidence Table 17). In the largest trial,40 with more than 200 patients in each arm, the RR of death or BPD at 36 weeks PMA in the iNO group was 0.97 (0.88, 1.07). A similar lack of benefit was found in a small trial of infants with birth weight > 1500 g when death was measured before hospital discharge or at 365 days for infants still hospitalized (RR 0.83 (0.43 1.62)).39 However, two other studies reported iNO was associated with a decreased risk of death or BPD at 36 weeks PMA. In Schreiber, 2003,58 with more than 100 infants in each arm, the risk of death or BPD was decreased 23 percent in the iNO treated group (RR 0.77 (0.60, 0.98)),and in a small study with 20 infants in each group, Dani, 200667 reported an 89 percent decrease in risk (OR 0.11 (0.02, 0.61)) among the iNO treated group (Figure 18).

The opposite direction of the effect of iNO therapy at 10 ppm in the two largest trials may be accounted for by the degree of illness of infants at study entry. Infants in the study of Schreiber, 200358 had a substantially lower oxygenation index at study entry (OI median (IQR) 7.3 (4.1, 12.3) iNO group versus 6.8 (4.4, 12.7) control group) than infants in the study of Van Meurs, 200540 ( OI mean (SD), 23 (17) iNO group versus 22 (17) placebo group) (see discussion of OI under Key Question 4) (Appendix E, Evidence Table 17). A meta-analysis revealed no significant difference in the risk of the combined outcome of death or BPD at 36 weeks PMA for infants treated with iNO at 10 ppm compared to controls, RR 0.81 (0.64, 1.03) (Figure 18).

In the four studies34, 60, 63, 64 that dosed iNO at 20 ppm or titrated to response, 50 percent to 100 percent of infants in each arm died or had BPD at 36 weeks PMA (Appendix E, Evidence Table 17). The pooled relative risk, RR 0.94 (0.84, 1.06), confirmed no difference to iNO treatment under these treatment protocols (Figure 18).

Neurodevelopmental impairment. In a followup study of Kinsella, 2006, Watson 200936 reported the rate of neurodevelopmental impairment (CP, severe hearing loss, blindness, or MDI or PDI < 70) at one year of age, corrected for gestational age at birth, was similar in the infants that received iNO at 5 ppm (35.4 percent) and those that received standard therapy (33.5 percent) groups (Appendix E, Evidence Table 17).

Only two of the studies giving iNO at 10 ppm evaluated long term neurodevelopmental outcome and they found opposite effects.40, 58 Reporting the followup of Van Meurs, 2005, Hintz, 2007,30 found no improvement in the combined outcome of death or neurodevelopmental impairment, defined as moderate to severe CP, blindness, deafness or an MDI or PDI < 70, at 18 to 22 months of age for infants treated with iNO compared to placebo, RR 1.07 (0.95, 1.19), or in death or moderate to severe CP, RR 1.17 (0.99, 1.38). Moderate to severe CP was increased in the iNO group (20 percent versus 11 percent), a difference that was not significant in univariate analysis, but reached significance in multivariable models after adjustment for infant characteristics at study entry in one model, RR 2.01 (1.01, 3.98), and infant characteristics and NICU morbidities in another model, RR 2.41 (1.01, 5.75). In both models the confidence intervals are wide (Appendix E, Evidence Table 17).

However, Mestan, 2005,56 reporting on the outcome of survivors at two years of age of the Schreiber, 2003 study, described a 47 percent decrease in the risk of cognitive impairment, defined as an a MDI < 70 (RR 0.53 (0.29, 0.94)), but no effect on motor impairment, defined as a PDI < 70 (RR 0.73 (0.33, 1.61)). Fewer infants treated with iNO had neurodevelopmental impairment, a composite variable including CP, blindness, hearing loss, and developmental delay, than infants treated with placebo (24 percent versus 46 percent respectively, p-value = 0.01). This difference in the composite outcome was the result of fewer infants with cognitive impairment in the iNO group as there was no difference between the groups in the rate of CP, vision or hearing loss (Appendix E, Evidence Table 17).

Three of the RCTs that gave iNO at 20 ppm or titrated the dose based on response reported long term developmental followup.35, 57, 76 Walsh, 201057 found no improvement in function at two years of age among infants treated with iNO in the study of Ballard, 2006.34 Rates of CP, cognitive delay, vision impairment, and hearing impairment were similar between the groups. In 30 month followup of infants enrolled in the Subhedar, 1997,64 and Bennett, 200176 studies similar rates of neurodevelopmental delay (MDI or PDI < 85), severe disability (defined as MDI or PDI < 70, or hearing loss or blindness), or death and severe disability were found between iNO treated infants and those receiving standard care. Huddy, 200835 found no difference at four to five years of age between the groups enrolled in the study of Field, 200563 in the rates of CP, cognitive delay, vision, and hearing impairment, combined moderate to severe disability, or in those free of impairment (23 percent iNO versus 19 percent placebo) (Appendix E, Evidence Table 17).

No meta-analyses were conducted for dose of iNO and neurodevelopmental impairment as the definition of impairment varied significantly between studies.

In summary, a meta-analysis of studies of 10 ppm dose of iNO found a statistically significantly reduction in BPD at 36 weeks PMA but meta-analyses found no statistically significant effect on death, or the combined outcome of death or BPD compared to control. The finding of a statistically significant effect at 10 ppm may be spurious as there is no clinical rationale for why that dose would be different from the others.34, 63-66 Results for neurodevelopmental impairment at this dose were inconsistent. There was no statistically significant different between iNO and control at doses of 5 ppm, 20 ppm or titrating the dose to the patients’ response.37

iNO with concurrent therapies. Two studies directly addressed the effect of iNO with concurrent therapies. In a factorial design, Subhedar, 199764 randomized 20 infants to treatment with iNO and dexamethasone and compared their outcome to 22 infants randomized to dexamethasone alone. Dexamethasone was given intravenously every 12 hours for six days at 0.5 mg/kg/dose for six doses, and 0.25 mg/kg/dose for the remaining six doses. All infants were less than 32 weeks gestational age at birth (Appendix E, Evidence Table 17). There was no difference between groups in the risk of death (RR 1.57 (0.76, 3.38)); BPD at 36 weeks PMA, defined as an oxygen requirement beyond 36 weeks PMA and an abnormal chest radiograph (RR 0.79 (0.44, 1.33)); the combined outcome of death or BPD at 36 weeks PMA (RR 1.05 (0.84, 1.25)); or BPD in survivors (RR 1.07 (0.71, 1.37)) (Appendix E, Evidence Table 17).

In a post hoc analysis, Ballard, 2006 34 reported that there were no significant differences in response to iNO by exposure to postnatal corticosteroids, though data were not shown. In two year followup, Walsh 201057 found that there was no interaction between iNO treatment and postnatal dexamethasone therapy given > three days after study enrollment compared with dexamethasone given ≤ three days from enrollment in a multivariable model of NDI.

Conclusion

Only two of the 14 randomized controlled trials that reported death, BPD, death or BPD, or neurodevelopmental impairment planned a priori to evaluate infants by subgroups,58, 64 so the evidence to answer this Key Question is not optimal. There is insufficient evidence to support treatment with iNO for acute lung disease within the first three days after birth. The one trial that compared infants treated at ≤ three days with those treated at four to 28 days found no difference between the groups. Given that none of the 14 RCTs reviewed in Key Question 1 found a significant difference in mortality or survival between those treated with iNO controls, many of which initiated therapy at less than three days, it is likely that early treatment is not beneficial. However, the duration of exposure to iNO varied in the trials and has yet to be systematically studied. For infants with developing BPD, earlier treatment (7 to 14 days) may prove to be more beneficial than later treatment (15 to 21 days) as shown by improved survival without BPD in the Ballard 2006 trial; an RCT with this primary hypothesis needs to be done. It is not surprising that delivery of iNO by high frequency ventilation conferred no convincing benefit, as high frequency ventilation alone has not been shown to reduce mortality, BPD, or improve neurodevelopmental impairment in preterm infants compared to conventional ventilation in systematic review.94 The optimal dose of iNO has yet to be determined. Our meta-analysis found a statistically significant effect for 10 ppm for BPD at 36 weeks PMA, but this may be a spurious finding. A similar effect was not seen for death or the composite outcome of death or BPD. Dose may be34, 37, 63-66 less important than the duration of iNO exposure, but data are insufficient to make that determination. The effect of concurrent therapies other than postnatal dexamethasone and iNO administration has not been studied in preterm infants.