Effect of nitric oxide on oxygenation and mortality in acute lung injury: systematic review and meta-analysisBMJ 2007; 334 doi: http://dx.doi.org/10.1136/bmj.39139.716794.55 (Published 12 April 2007) Cite this as: BMJ 2007;334:779
- Neill K J Adhikari, lecturer1,
- Karen E A Burns, assistant professor1,
- Jan O Friedrich, assistant professor1,
- John T Granton, associate professor1,
- Deborah J Cook, professor2,
- Maureen O Meade, associate professor2
- 1Interdepartmental Division of Critical Care Medicine, University of Toronto, Toronto, Canada
- 2Departments of Medicine and Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, Canada
- Correspondence to: N K J Adhikari
- Accepted 23 January 2007
Objective To review the literature on the use of inhaled nitric oxide to treat acute lung injury/acute respiratory distress syndrome (ALI/ARDS) and to summarise the effects of nitric oxide, compared with placebo or usual care without nitric oxide, in adults and children with ALI or ARDS.
Design Systematic review and meta-analysis.
Data sources Medline, CINAHL, Embase, and CENTRAL (to October 2006), proceedings from four conferences, and additional information from authors of 10 trials.
Review methods Two reviewers independently selected parallel group randomised controlled trials comparing nitric oxide with control and extracted data related to study methods, clinical and physiological outcomes, and adverse events.
Main outcome measures Mortality, duration of ventilation, oxygenation, pulmonary arterial pressure, adverse events.
Results 12 trials randomly assigning 1237 patients met inclusion criteria. Overall methodological quality was good. Using random effects models, we found no significant effect of nitric oxide on hospital mortality (risk ratio 1.10, 95% confidence interval 0.94 to 1.30), duration of ventilation, or ventilator-free days. On day one of treatment, nitric oxide increased the ratio of partial pressure of oxygen to fraction of inspired oxygen (PaO2/FiO2 ratio) (13%, 4% to 23%) and decreased the oxygenation index (14%, 2% to 25%). Some evidence suggested that improvements in oxygenation persisted until day four. There was no effect on mean pulmonary arterial pressure. Patients receiving nitric oxide had an increased risk of developing renal dysfunction (1.50, 1.11 to 2.02).
Conclusions Nitric oxide is associated with limited improvement in oxygenation in patients with ALI or ARDS but confers no mortality benefit and may cause harm. We do not recommend its routine use in these severely ill patients.
Acute lung injury (ALI) and acute respiratory distress syndrome (ARDS), defined by acute hypoxaemia and bilateral lung infiltrates on radiography without left atrial hypertension,1 are characterised by inflammation of the alveolar-capillary membrane triggered by various insults.2 Because the pathophysiology involves mismatching of ventilation and perfusion and pulmonary hypertension, the possibility of using inhaled nitric oxide (NO) generated considerable interest.3 Nitric oxide is a selective pulmonary vasodilator and has anti-inflammatory properties.4 5 Based on limited data on efficacy, clinicians rapidly adopted this therapy; 63% of European intensive care specialists surveyed in 1997 reported using it, primarily for ALI or ARDS.6 A more recent survey of specialists in Ontario, Canada, found that a substantial proportion (39%) reported using nitric oxide at least sometimes in selected patients with ARDS.7
A systematic review and meta-analysis of nitric oxide published in 20038 9 that included five randomised controlled trialsw3-w7 found no effect on mortality or ventilator-free days; one trial showed improved oxygenation.w3 Because confidence intervals were wide, the authors concluded that the effects of nitric oxide on morbidity and mortality were uncertain. We have incorporated data from new randomised controlled trials to evaluate the effects of nitric oxide on pulmonary physiology (oxygenation and pulmonary arterial pressure) and important clinical outcomes (mortality, duration of ventilation, and adverse effects) in patients with established ALI or ARDS.
We electronically searched Medline, CINAHL, Embase, and CENTRAL (to October 2006), limiting citations to randomised controlled trials. We also searched proceedings of four conferences (1994-2006), screened bibliographies of retrieved studies and recent review articles,10 11 12 13 14 15 16 17 18 and contacted content experts to identify additional trials. There were no language restrictions. Further details of the search strategy and other aspects of study methods are on bmj.com.
Two reviewers independently screened studies for inclusion, retrieved potentially relevant studies, and decided on study eligibility. We selected parallel group trials that enrolled adults or children (excluding neonates), with ≥80% of patients or a separately reported subgroup having ALI or ARDS (using authors' definitions). Included trials compared nitric oxide with placebo or usual treatment (not prevention) for ALI or ARDS and reported mortality (at any time), duration of ventilation, ventilator-free days, or pulmonary physiological parameters on days one to four of treatment (PaO2 (partial pressure of oxygen)/FiO2 (fraction of inspired oxygen); oxygenation index, defined as 100 × mean airway pressure/(PaO2/FiO2); mean pulmonary arterial pressure). We included trials with cointerventions applied equally in both groups. We assessed agreement between reviewers for trial eligibility using Cohen's κ.19
Data abstraction and validity assessment
Two reviewers independently abstracted data and methods from included trials. We resolved by consensus any disagreements that remained after contacting trial authors. From included studies we abstracted method of randomisation and allocation concealment, blinding of caregivers and outcomes assessors, and number of withdrawals after randomisation and determined whether mechanical ventilation, weaning, and sedation were standardised or applied equally in treatment groups.
We attempted to contact authors of all included trials to request additional data and clarify data and methods if necessary.
Quantitative data synthesis
Our primary outcome was mortality in hospital (or, if not available, mortality in the intensive care unit or at 28 or 30 days). We decided a priori to combine trials with less than half of patients crossing over from control to nitric oxide arms in analyses of clinical outcomes. Our analyses adhered to the intention to treat principle. In studies with two or more nitric oxide groups receiving different doses, we combined data to determine an overall effect for the nitric oxide group.
Secondary outcomes included duration of ventilation, ventilator-free days to 28 or 30 days, and pulmonary physiology. We decided post hoc to combine data on renal dysfunction after obtaining outcomes for most randomised patients, but we describe other adverse events qualitatively.
We used random effects models20 implemented in Review Manager 4.2.7 (Cochrane Collaboration, Oxford) for all analyses and considered P≤0.05 (two sided) as significant. We report binary outcomes as risk ratios and continuous outcomes as weighted mean differences (measure of absolute change) and ratios of means (measure of relative change).21 Summary effect estimates are presented with 95% confidence intervals.
We assessed homogeneity between studies for each outcome using the Cochran Q statistic,22 with P≤0.10 indicating significant heterogeneity,23 and I2 24 25 with suggested thresholds for low (25%-49%), moderate (50%-74%), and high (≥75%) values. We developed several a priori hypotheses to explain significant heterogeneity (excluding duration of ventilation and ventilator-free days), including dose and duration of nitric oxide therapy and whether therapy was restricted to patients whose oxygenation improved acutely (“nitric oxide responders”) or to those with ARDS (the more hypoxaemic subset of ALI).
Electronic database searches yielded 1262 citations. After evaluating these citations, conference abstracts, review articles, and bibliographies of included trials, we included 12 parallel group randomised controlled trialsw1-w12 (fig 1)⇓. The two reviewers completely agreed (κ=1) on the selection of included studies. We obtained additional information from 10 authors (new clinicalw1 w2w5w7 w8w11 or physiological dataw1 w2w7w11; clarifications of dataw6w9 or methodsw1-w3w5-w11).
Study characteristics and methodological quality
Table 1⇓ describes the included studies, two of which were published as abstracts only.w2w7 Data from one trial were distributed in two abstracts,w8w13 and data from another trial were distributed in two articles.w6w14 Trials randomised 1237 patients (median 40; range 14-385) with ALI or ARDS. Two trials enrolled only children,w1w6 one trial included a few children,w4 and the remaining trials enrolled only adults. All patients met American-European Consensus Conference1 oxygenation criteria for ARDS except for one trial that included some patients with ALI.w12 Seven trials used a fixed dose of nitric oxide (median 10 ppm; range 5-10 ppm),w1 w2w6w8w10-w12 and five used the lowest dose to achieve an oxygenation responsew4 w5w7w9 or randomised patients to different doses.w3 One trial enrolled only patients whose oxygenation improved after a nitric oxide challenge (“nitric oxide responders”),w7 and one used a cointervention (a recruitment manoeuvre) in both groups.w11 Trials continued nitric oxide until prespecified gas exchange end pointsw1w3w5w7-w10w12 or for a fixed period of time after which nitric oxide was tapered by using gas exchange criteriaw4w6 or managed at clinicians' discretion.w2 One trial did not report on criteria for stopping nitric oxide.w11 The median duration of administration was 6.5 days (range 3.5-9.0 days; data available from five trialsw5w7-w9w11). One trial randomised patients to nitric oxide or control for 24 hours, after which all patients received nitric oxide.w1 In five other trials, control patients received nitric oxide as rescue therapy after randomisation if they met prespecified criteria (<50% of controls in three trialsw4w7 w8 and ≥50% in two trialsw2w6). Not for profit agencies funded five trials,w1w4-w6w10 industry funded two trials,w3w12 both sources funded or supported four trials,w2w7-w9 and one trial did not report this information.w11
The 12 trials had good scientific quality (table 2)⇓. Ten concealed randomisation,w1-w3w5-w8w10-w12 and five blinded clinicians.w2 w3w6w8w12 Mechanical ventilation was delivered according to protocol in three unblinded trialsw5w10 w11 and one blinded trialw2 and according to guidelines in three blinded trials.w3w6w12 Six trials described or standardised at least one other cointervention, such as corticosteroids,w3 sedation,w5 prone ventilation,w5w9-w12 and ventilator weaning.w11 w12 All trials had complete follow-up, analysed patients by assigned group, and withdrew no one from clinical outcomes analyses. One trial stopped early because of slow enrolment (achieving 45% of the planned sample sizew7), and another trial enrolled 75% of the planned sample, for unclear reasons.w12
Effect of nitric oxide on clinical outcomes
We combined nine trialsw3-w5w7-w12 in the mortality analysis (three were placebo controlledw3w8w12; five used “usual care” controlsw4 w5w7w9 w10; one used recruitment manoeuvres in both armsw11). We combined three trials that reported duration of ventilation (including all patientsw10 w11 or only survivorsw7) and five trials reporting ventilator-free days.w3w5w8w11 w12
Meta-analyses (table 3)⇓ showed that nitric oxide did not affect mortality (risk ratio 1.10; 95% confidence interval 0.94 to 1.30; fig 2)⇓, duration of ventilation (17% increase, −20% to 70%; 3.6 additional days, −4.0 to 11.1 days), or ventilator-free days (6% decrease, −16% to 6%; 0.6 fewer days, −1.8 to 0.7 days). There was moderate to high heterogeneity between studies for duration of ventilation only.
A funnel plot of standard error versus risk ratio for mortality did not suggest publication bias (fig 3)⇓.
Effect of nitric oxide on physiological outcomes
On the first day of therapy, NO was associated with small improvements in the PaO2/FiO2 ratio (nine trials; 13% higher, 4% to 23%; 16 mm Hg higher, 4 mm Hg to 27 mm Hg; fig 4)⇓ and oxygenation index (three trials; 14% lower, 2% to 25%; 3 cm H2O/mm Hg lower, 0.5 cm H2O/mm Hg to 5 cm H2O/mm Hg; fig 5).⇓ Some evidence suggested that improvements in oxygenation in the nitric oxide group persisted beyond day one. The PaO2/FiO2 ratio was higher on day two and four (but not on day three, and only in the ratio of means analysis on day two). The oxygenation index remained lower on days two, three, and four (only in the weighted mean difference analysis on day three), but only onew3 (days two and four) or twow3w6 (day three) trials contributed data. Differences in mean pulmonary arterial pressure were not significant on any day.
There was no evidence of important statistical heterogeneity in the physiological outcomes.
Table 4 gives details of adverse effects.⇓ All 12 trials gave information about methaemoglobin concentrations. Four nitric oxide patients (of 651 randomised) and three control patients (of 586 randomised) developed >5% methaemoglobinaemia.w3w7w12 One trial reported three patients developing raised nitrogen dioxide concentrations; all had received 80 ppm nitric oxide .w3 Nitric oxide increased the risk of renal dysfunction in one unblindedw7 and three blindedw3w8w12 trials that enrolled 72% of patients in all included trials (risk ratio 1.50, 1.11 to 2.02; fig 6)⇓. Other adverse events were variably reported, and we did not combine these data.
The routine use of inhaled nitric oxide is not beneficial for patients with acute lung injury (ALI) and acute respiratory distress syndrome (ARDS). Our meta-analysis included 12 trials that randomly assigned 1237 patients and investigated the effects of inhaled nitric oxide in such patients. We found no benefit of nitric oxide on survival and an increased risk of renal dysfunction. Oxygenation improved over the first 24 hours (13% relative increase in PaO2/FiO2 ratio; 14% decrease in oxygenation index), with some data suggesting improvements to 96 hours. Given the limited physiological improvements and possible harm, we cannot recommend routine use of nitric oxide in these patients.
The trend towards increased mortality in patients receiving nitric oxide was highly consistent across trials, with no trial dominating the meta-analysis. Given the strength and magnitude of this trend, consistency across trials, biological plausibility,18w10 and the finding of other potential adverse effects of nitric oxide (for example, renal failure), our analysis raises concerns about its nitric oxide in this setting.
Descriptive analyses suggest that methemoglobinaemia and raised nitrogen dioxide concentration are not common or clinically important consequences, except possibly in patients receiving high doses (at least 80 ppm) of nitric oxide for several days. Data from four large trials representing nearly three quarters of all randomised patients showed an increased risk of renal dysfunction in patients receiving nitric oxide. Cautious interpretation is warranted, however, as this result was a post hoc analysis and is potentially subject to publication bias (we were unable to obtain explicit data on renal outcomes in eight of 12 smaller trials, in which this relation may not have been measured or observed). In addition, the potential physiological mechanisms linking administration of inhaled nitric oxide to acute renal dysfunction—inhibition of mitochondrial and enzymatic function and damage to deoxyribonucleic acid and membranes—are controversial because of its simultaneous protective effects on renal blood flow and leukocyte adhesion.26
Why nitric oxide may not be beneficial
There are several possible explanations for the lack of benefit of routine administration of nitric oxide in patients with ALI/ARDS. Firstly, short term physiological improvements in oxygenation seem to have no impact on patients' survival,27 possibly because oxygenation is not necessarily related to severity of lung injury. Secondly, as most patients with ARDS die of multiple organ failure rather than refractory hypoxaemia,28 small changes in oxygenation might not lead to improvements in outcome. Thirdly, the prolonged fixed dosing regimen in most trials may have attenuated benefit over time because of increased sensitisation, dampening the oxygenation benefit while continuing to expose patients to toxic effects such as oxidative damage.18w10 Fourthly, the benefits of nitric oxide may have been overwhelmed by a harmful mechanical ventilation strategy, which perpetuated multiple organ failure.29 This, however, would not account for our finding of potential harm. Finally, trials restricting enrolment to patients with an acute oxygenation response to nitric oxide may have found a positive effect on mortality, although this hypothesis was not supported in one trial.w7
Strengths and limitations
We used several methods to reduce bias (comprehensive literature search, duplicate data abstraction, prespecified criteria for methodological assessment and analysis) and analysed a comprehensive set of clinical and physiological outcomes. We were unable to obtain anyw4w12 or completew3 additional information from three trials. Considering secondary clinical outcomes, we expected to find variation between trials in duration of ventilation and ventilator-free days related to different populations of patients. We analysed these outcomes, while acknowledging the limited interpretability of this analysis. Finally, given the small number of trials contributing to analyses of many physiological outcomes, the tests for heterogeneity were underpowered.
Although our results do not exclude the possibility that some subgroups of patients may benefit from nitric oxide, the consistent lack of a mortality benefit across trials mitigates this possibility. The included trials did not specifically study the issue of nitric oxide as rescue therapy for patients with critically low oxygenation. With nitric oxide, short term improved oxygenation in these patients may create a window for other strategies to improve lung function, such as treatment of the underlying cause of ARDS.
A previous systematic review and meta-analysis of inhaled nitric oxide for acute hypoxaemic respiratory failure8 9 included fewer randomised controlled trialsw3-w7 and found no effect on mortality (risk ratio 0.98, 95% confidence interval 0.66 to 1.44; two trials, 204 patients). Our report is consistent with this work and extends it by including more trials, thus narrowing the confidence limits around the estimate of mortality. We also provide new estimates of the impact of nitric oxide on other clinical and physiological end points and raise the possibility of harm induced by nitric oxide.
In conclusion, our systematic review and meta-analysis found that inhaled nitric oxide improved oxygenation in patients with ALI and ARDS at 24 hours of therapy, with some evidence for a more prolonged effect. Given that the best available evidence suggests no survival advantage and possible increased mortality and renal dysfunction with nitric oxide, we do not recommend its routine use. Despite a lack of evidence for benefit, some clinicians may still consider nitric oxide for life threatening hypoxaemia, in conjunction with other supportive therapies. Given the challenges of enrolling such severely ill patients into large trials, definitive data supporting or refuting a role for nitric oxide in such desperate situations may not be forthcoming, leaving clinicians to rely on their judgment and the current evidence.
What is already known on this topic
Inhaled nitric oxide continues to be used to improve oxygenation in patients with acute lung injury, despite no clear supporting evidence and several discouraging reviews and editorials
A previous meta-analysis in 2003 included five randomised trials of nitric oxide; there are now 12 trials
What this study adds
Nitric oxide improves oxygenation temporarily but does not improve survival and may cause harm
We do not recommend routine use of nitric oxide in patients with acute lung injury
We thank Phil Dellinger, Emily Dobyns, Herwig Gerlach, and Sangeeta Mehta for providing additional information about their trials; Pascal Beuret, Gilbert Blaise, Ronald Day, Stefan Lundin, Kwang Joo Park, Didier Payen, and Benoît Vallet for providing additional outcomes data; Natasha Stankovic for assistance in translation; and Jim Julian for constructive comments on an earlier draft of the manuscript.
Contributors: NKJA conceived and designed the study, acquired data, analysed and interpreted data, and drafted the manuscript. KEAB contributed to study design and acquired and interpreted data. JOF acquired and interpreted data. JTG interpreted data. DJC and MOM contributed to study design and interpreted data. All authors revised the manuscript for important intellectual content and approved the final version. NKJA is guarantor.
Competing interests: None declared.
Ethical approval: Not required.