INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Initially viewed solely as a surfactant abnormality, identical to that seen in neonatal respiratory distress syndrome (1), acute respiratory distress syndrome (ARDS) in adults is now considered as an inflammatory disorder that transcends pulmonary lesions and includes involvement of the microvasculature in multiple organ systems (2). It is characterized by diffuse radiographic pulmonary infiltrates because of an increased alveolocapillary permeability (3).

The pulmonary lesion is mainly due to an inflammatory interaction between platelets, leukocytes, mononuclear cells, macrophages, and endothelial cells, where oxidative stress is prominent (4). The clinical signs include refractory hypoxemia, mainly because of ventilation/perfusion mismatching, and important right-to-left shunting, high pulmonary arterial pressure often associated with right ventricular dysfunction, noncardiogenic pulmonary edema, and decreased lung compliance with few recruitable lung units (3). If the landmark work of Murray and colleagues (3) helped establish a universal definition of ARDS, the report of the American-European consensus conferences on ARDS (2) outlined the heterogeneity of populations and nonuniformity of therapy, therefore making it difficult to identify outcome predictors.

Nitric oxide (NO) is a vasodilator whose main mechanism of action is the stimulation of soluble guanylate cyclase and the production of cyclic guanosine monophosphate. It also inhibits platelet (5) and leukocyte (6) activation and adhesion and their production of active intermediates (7). As previously demonstrated, local effects of inhaled NO (inhNO) on pulmonary hypertension and consequently right ventricular dysfunction, oxygenation (8), inflammation (9), or pulmonary edema (10) argue for its use in ARDS. Moreover, the extrapulmonary effects of inhNO, i.e., platelet activation (11) and cardiac (12) and renal (13) protective effects could also be helpful for the treatment of patients with ARDS. All these effects are likely to decrease ARDS mortality.

Since the first enthusiastic reports of Rossaint and colleagues (8), and Gerlach and coworkers (14) (respective survival rate of 80 and 100%), and despite the recommendations of the American-European ARDS consensus (2), no study has been appropriately designed to address the clinical outcome of patients with ARDS treated with inhNO. Most of the previous reports on the use of inhNO in ARDS were small case series or crossover designs comparing different therapeutics (8, 12, 14). Results from both types of studies were very limited because of selection bias and the absence of a control group followed in parallel. Crossover design also has some major limitations when the condition is not stable in time, and death is frequent. To our knowledge, this is the first report of a randomized controlled trial of patients with ARDS treated with inhNO that addresses its efficacy, not only on gas exchange but also on mortality. The objectives of the trial were to determine: (1) the efficacy of inhNO on lung function; (2) the impact of inhNO on morbidity and mortality; (3) the chances of success and necessary conditions for implementing a larger multicenter trial.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Protocol

This study was approved by the Human Research and Ethics Committee of the Centre Hospitalier de l'Université de Montréal, and was in accordance with the 1975 Helsinki Declaration. From October 1994 to August 1995, all patients 18 to 75 yr of age suffering from ARDS (the lung injury score [LIS]  2.5 of Murray and colleagues) hospitalized in the medical or surgical intensive care units (ICU) were included in the study. Pregnant women, patients with cardiogenic pulmonary edema (pulmonary capillary wedge pressure > 18 mm Hg), or with severe immunodepression from end-stage neoplasia were excluded. Written informed consent was obtained from a close relative. Patients were then randomized to receive either usual care (control group) or usual care plus inhNO (experimental group).

Severity scoring systems (APACHE II [18] and LIS) were calculated for each patient at randomization (Day 0). Patients in both groups were mechanically ventilated following a standardized protocol applied for the fraction of inspired oxygen (FI_O2), tidal volume (VT), ventilatory mode, and positive end-expiratory pressure (PEEP). Sedation, curarization, intravenous perfusion, blood transfusion, parenteral or enteral feeding were also standardized (clinical protocols are available upon request from us). All patients had arterial lines and pulmonary arterial catheters in place for continuous monitoring and blood sampling purposes. Pulse oximetry, end-tidal carbon dioxide tension (PET_CO2) were also continuously monitored. Arterial blood gases and pH and pulmonary and systemic hemodynamic parameters were measured at least once every 8 h, whereas venous blood gases, methemoglobinemia, cardiac output, and cardiopulmonary profiles were obtained at least daily.

Therapeutic success was defined as successful weaning from mechanical ventilation with extubation within 30 d after inclusion in the study. Briefly, all patients, whether ventilated with pressure-controlled or conventional ventilatory volumetric SIMV (synchronized intermittent mandatory ventilation) mode were weaned with progressive decrease in SIMV rate supplemented with pressure support. Afterwards, if patients maintained the usual clinical parameters with pressure support  10 cm H₂O, PEEP  5 cm H₂O, and FI_O2  0.5 they were extubated. This protocol was applied in a similar manner to all patients in the study. Therapeutic failure was defined as death before 30 d or pursuit of mechanical ventilation (patient always intubated) from Day 31 onwards. Lung function status was assessed by the hypoxia score (HS = Pa_O2/FI_O2), alveolar dead space (VDA/VT = 1  [PET_CO2/Pa_CO2]), lung compliance (CL = VT/[Ppeak  totalPEEP]), and venous admixture [VA/T = 100 × (CC_O2  Ca_O2)/(CC_O2  Cv_O2)] (Pa_O2, Pa_CO2: partial pressures of arterial oxygen and carbon dioxide; Ppeak: peak airway pressure; CC_O2, Ca_O2, and Cv_O2: capillary, arterial, and venous oxygen contents). Nonresponders to inhNO were defined as patients presenting a  20% increase in HS after initial optimal inhNO administration. All patients, responders and nonresponders, were included in the statistical analysis.

Inhaled NO Administration

Cylinders contained 900 ppm of NO in N₂ gas (Air Liquide Ltée., Montréal, PQ, Canada) with NO₂ < 5 ppm. The NO/N₂ mixture was cyclically injected into the inspiratory line of the mechanical ventilator (7200; Puritan-Bennett Inc., Carlsbad, CA) using an inhNO administration system developed at our institution (19). The fractions of inspired NO (FI_NO) and NO₂ (FI_NO2) delivered to the patients were measured electrochemically (Polytron NO, NO₂; Dräger AG, Lübeck, Germany). The calibration of the electrochemical device was performed at least weekly, and validated with chemiluminescence (CLD 700 AL NO/NOx analyzer; Ecophysics Tecan AG, Dürten, Switzerland).

Optimal FI_NO Determination

In the experimental group, the lowest FI_NO that produced the greatest improvement in Pa_O2 was determined every day (optimal FI_NO = minimal dose with maximal efficacy on Pa_O2). During the determination of optimal FI_NO, no treatment or procedure susceptible of changing the arterial oxygenation or the hemodynamic status (defined as inhNO evaluative parameters) was allowed.

Initial Optimal FI_NO Determination

The initial FI_NO was 2.5 ppm. The parameters were collected immediately before (baseline values) and 10 min after introduction of inhNO. The value of FI_NO was then increased stepwise according to the following pattern of concentrations: 5, 10, 20, 30, and up to a maximum of 40 ppm. The decision to increase FI_NO was directed by the observed changes in Pa_O2 assessed 10 min after the modification of FI_NO. Basically, FI_NO was increased until the increase in Pa_O2 was smaller than 5% from the previous assessment. Theoretically, if there was no beneficial effect with all doses tested (2.5 to 40 ppm), FI_NO was kept at 5 ppm, but this situation never occurred. Once determined, initial optimal FI_NO was kept unchanged for 24 h.

Daily Optimal FI_NO Determination

Every morning FI_NO was decreased stepwise following the same pattern; below 2.5 ppm the two stages were 1 and 0.5 ppm, down to 0 ppm if possible. Similarly, Pa_O2 was collected 10 min after each modification, and the decrease in FI_NO was stopped when it induced a larger than 5% decrease in Pa_O2. In that case the previous FI_NO was kept for 24 h. If the first decrease in FI_NO induced a fall in Pa_O2 by more than 5%, effects of higher FI_NO were determined stepwise, as previously described.

Inhaled NO Weaning

In the eventuality that no beneficial effects with inhNO was observed for the first 2 d, we had anticipated to withdraw the therapy. Daily optimal FI_NO determination, through regular reverse dose-response assessments, allowed patients to be gradually weaned from inhNO. The patients had the criteria to be weaned from inhNO treatment if they maintained a Pa_O2  70 mm Hg with FI_O2  0.4 and PEEP  8 cm H₂O during the daily optimal FI_NO determination at 0 ppm FI_NO for two consecutive days.

Inhaled NO Reintroduction

After successful weaning, treatment was resumed if the pulmonary status deteriorated. Criteria for reintroduction were: FI_O2 > 0.4, PEEP > 10 cm H₂O, and Pa_O2  60 mm Hg. Initial optimal FI_NO was determined as previously described.

Statistical Analysis

For the comparison between groups, we distinguished two phases to separate the initial effect of inhNO (8, 20) from its remaining effect during prolonged administration. The initial response to inhNO was defined as the difference between the baseline value and the value measured under initial optimal FI_NO. The follow-up effect was assessed by the rate of change from Day 1 to the last day of follow-up. Differences in lung function were determined first by examining the appropriateness of fitting linear regression slope for each individual patient, second by estimating the linear slope for each individual patient, third by weighting the individual slopes according to the precision with which they were estimated, and finally by comparing the difference between the inhNO and control group mean slopes using Student's t test (SAS for Windows^TM, release 6.11; SAS Institute Inc., Cary, NC). The appropriateness of the linear approximation was determined by adding a quadratic term to the model for an individual patient's lung function and examining whether the addition provided a statistically significant improvement in fit. Because patients could contribute different numbers of observations to the estimation of their respective slopes, not all individual slopes were estimated with the same precision. In order to account for this, individual slopes were weighted inversely proportional to the square of standard error (1/ SE²) when they were combined to obtain an average slope. The therapeutic success rates in each group were compared using a log-rank test. For the primary analysis, an intention to treat approach was used. Those patients who died were assigned the maximum possible length of ventilation, 30 d. Subsequent analysis examined outcomes in the subgroups of patients who remained alive for the duration of the follow-up. Data are expressed as mean ± SE.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients

Thirty patients with ARDS were randomized, 15 to each of the experimental and control groups. Established ARDS was present in 13 medical and 17 surgical ICU patients. No trauma-induced ARDS was noted during the study period. Twelve patients in the control group, and 13 in the experimental group were septic, and all others recruited resulted from direct lung injury (infectious pneumonia, aspiration, or toxic inhalation). The two groups were similar with regard to most baseline demographic and prognostic variables (Table 1). patients treated with inhNO, however, were on average slightly more sick, with a HS that was lower, and an APACHE II score that was higher, than patients in the control group.

Initial Effects of inhNO

The initial optimal FI_NO decreased pulmonary arterial pressure without changing systemic arterial pressure and heart rate (Table 2). Inhaled NO significantly improved the gas exchange parameters: increases in Pa_O2 by 66.2% (p = 0.0025) and H⁺ concentration by 4.5% (p = 0.0025); decreases in Pa_CO2 by 7% (p = 0.0001) and HCO₃ by 2.8% (p = 0.02). Although PET_CO2 remained unchanged, VDA/VT and AaPO₂ decreased significantly. The mean initial optimal FI_NO established for the first day was 8.5 ± 2.6 ppm. five of 15 patients were nonresponders to inhNO (increase in HS  20%). There was no difference in the initial mean HS between responders and nonresponders (127.7 ± 25.6 versus 118 ± 19.6 mm Hg, respectively). Comparison between groups showed that the initial changes of Pa_O2, HS and VA/T were significantly larger in the experimental group than in the control group (Table 3).

Follow-up Effects

Excluding the initial effect, we did not observe any difference between groups. There was not sufficient evidence to support the use of a quadratic model for any of the measures of lung function considered (the proportion of patients with significant quadratic terms never exceeded 20%). Accordingly, for each patient we characterized the change in lung function over the follow-up period by the slope of the corresponding linear regression of lung function on time. As shown in Table 3, the mean weighted linear slopes from Day 1 onward for PEEP, CL, and VA/T were not different between groups; there was some evidence (p = 0.06) of a difference in the change of HS, with a tendency of greater decline in the inhNO-treated group than in the control group. It should be noted that the estimated mean weighted linear slopes were sometimes highly variable between individual patients, e.g., for HS in the inhNO group, comprising essentially an equal number of positive and negative individual slopes. The mean FI_NO used during the study, based on optimal dosage determination, was 5.6 ± 1.8 ppm. The mean duration of inhNO treatment was 8.1 ± 1.3 d (range: 28 to 453 h), FI_NO2 was always  1 ppm, and excessive methemoglobinemia was never detected.

Survival Rate, Duration of Mechanical Ventilation, and Therapeutic Success Rate

The 30-d mortality and the duration of mechanical ventilation were almost identical in the two groups: nine of 15 deaths and 22.3 ± 2.5 d of ventilation in the inhNO group versus eight of 15 deaths and 24.1 ± 2.5 d in the control group. A majority of the deaths (11 of 17) occurred from multiple organ dysfunction syndrome (MODS). Although the subgroup of survivors treated with inhNO showed a 20% reduction in the duration of mechanical ventilation when compared with the control group (10.8 ± 1.2 versus 12.8 ± 4.2 d, respectively), the difference was not statistically significant (p = 0.44). The global evolution of therapeutic success was not different between the two groups (p = 0.83) (Figure 1). In the experimental group, the final therapeutic success rate was slightly higher (40%) and reached earlier (at Day 15) than in the control group (33.3% at Day 30). Of the five patients of both groups with direct ling-injury-induced ARDS only one died, in the control group. Consequently, the therapeutic success rate of the remaining patients with sepsis-induced ARDS was 28% (30.8% in the inhNO, and 25% in the control group). Four of five inhNO nonresponders died, and the therapeutic success rate in inhNO responders was 50%.

View larger version (10K): [in this window] [in a new window]   Figure 1.   Therapeutic success rate evolution of control and experimental (inhaled NO) groups. Therapeutic success was defined as successful weaning from mechanical ventilation with extubation within 30 d after inclusion in the study. There was no evidence of difference between the two groups (p = 0.83).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This is the first report of a randomized controlled trial of patients with ARDS treated with inhNO that addresses these issues. Small sample size, heterogeneous populations and treatments, as well as selection bias, plagued previous studies on the effect of inhNO in patients with ARDS. Moreover, none of them used a control group without inhNO followed in parallel except Chollet-Martin and colleagues (9) whose main objective was not the survival rate. Several of these studies carried out a crossover design with prostacyclin (8, 12, 15) or almitrine (16, 17) to study the effect of inhNO on hypoxemia. This design, however, fails to address important clinical outcomes such as survival or weaning and is very limited because of the ARDS instability in time. The lack of consistency in previous studies could result from the high variety of ARDS etiologies, the large FI_NO (10 ppb [20] to 128 [21] ppm) and age (1 [15] to 81 [22] yr) ranges. Most of the investigators used a unique FI_NO dose, which, in our opinion, does not optimize treatment. Finally, several studies used concomitant treatments such as ECMO (8, 12, 14, 20, 23), permissive hypercapnia (22, 24, 25), or vasoactive drugs (16, 21, 25, 26), which makes it difficult to compare the outcome. In our trial, the population was selected on strict criteria and received uniform standardized care for their conditions. Random allocation of the intervention as well as strict objective assessment of the parameters ensured absence of bias.

Our results confirm those of previous studies (15, 20, 26, 27) in so far as showing that inhNO initially exerts a moderate pulmonary vasodilatory effect that is accompanied by an improvement in alveolocapillary gas exchange and an absence of clinical systemic hemodynamic effects (Tables 2 and 3). The beneficial effect resulting in improved oxygenation is likely related to the redistribution of blood flow from unventilated shunted areas to ventilated but under perfused areas, the so- called "steal phenomena" (16, 26). Unlike that in a study of patients with chronic obstructive pulmonary disease (29), no negative response to introduction of inhNO was observed in our study. By dilating some constricted pulmonary vessels in ventilated lung areas, inhNO improved the high ratio ventilation/perfusion inequality, thereby reducing AaPO₂, and VDA/ VT. As a consequence, Pa_CO2 was slightly but significantly reduced. In our study, PET_CO2 was unchanged; other investigators have similarly found either no increase (23, 27) or only a slight (1 to 3 mm Hg) increase (21, 23) in this parameter. The effect of inhNO over time in comparison with a control group has never been published. The initial effects on HS and VA/ T (Table 3) were significantly greater in the experimental group than in the control group. However, after the initial period, there was no evidence of continued improvement in the experimental group or in the control group. There was no significant difference in duration of mechanical ventilation and mortality, and, consequently, the therapeutic success rate was similar in both groups. The slight trend of inhNO improvement through the duration of mechanical ventilation observed with the subgroup of survivors is difficult to interpret. It could represent a real effect or be a product of selection bias.

The contrast between a strong positive initial effect on lung function and the absence of effect on global mortality is striking and may yield different explanations. As noted above, a slight disadvantage on baseline severity scoring systems and lung status was observed in the experimental group. This difference, however, is not likely to explain the absence of survival difference because there was no correlation between the severity of initial scoring systems and the 30-d mortality. A multivariate model, adjusted for the initial severity score, did not change the results. Our results show that after the first day the two groups presented similar evolutions, suggesting that both groups reached the same maximal level of possible improvement, but that patients treated with inhNO did it more quickly. Therefore, inhNO did not change the general health status, which was already profoundly altered (mean APACHE II score = 27.4) when patients were included (established ARDS). Our results are compatible with the fact that the death during the acute phase of ARDS is usually attributable to the underlying illness or injury (30), whereas later deaths are mainly due to MODS (31), and they corroborate the comments of Wenstone and Wilkes (32) that new therapies directed solely at treating the lung will be unlikely to significantly reduce the mortality from what is essentially an inflammatory systemic disease.

In this context one may see the evolution of HS over time as the effect of optimal standard ventilation whose effectiveness is limited by the impaired lung condition; inhNO might help reach this maximum attainable level, without transforming the respiratory function or the final prognosis. It is possible that earlier treatment with inhNO could be more effective on mortality, and ARDS may be easier to reverse (33), or that inhNO administered to patients with direct lung-injury- induced ARDS (without MODS) may also have a positive effect on mortality as well. In the studies done by Rossaint and colleagues (8) and Gerlach and coworkers (14), the majority of patients had ARDS resulting from direct lung injury, and the survival rate with inhNO was high (11 of 13). In our study, the number of patients so affected was small (n = 5) and the prognosis was good in both groups: two of two, and two of three, respectively, in the inhNO and control groups. We also found that nonresponders to inhNO (initial increase in HS  20%) had a very poor prognosis (four of five deaths). All these patients, however, were treated because their HS improved by more than 5% with the therapy. The lack of a demonstrable effect on mortality does not imply that administration of inhNO is worthless (34). The rapid improvement in lung function in comparison with the control group is a positive element that may facilitate ventilatory support; hence, decrease iatrogenic pulmonary pathology. This aspect supports the observation made by Petros and colleagues (34) that mortality is not a good criterion to assess the role of a new therapeutic modality in the ICU, and that reducing morbidity is a more adequate achievable end point. Eventually, the additive effects of a number of therapies decreasing morbidity in patients with ARDS may lead to a decrease in mortality (25). It has been shown that the overall mortality has lessened in ARDS, although no single therapeutic approach improved it (35).

Our study confirms the initial efficacy of inhNO on lung function but does not show any significant difference between groups after the first day of follow-up. We observed a more rapid resolution of pulmonary changes in blood gas criteria in ARDS, which was not followed by beneficial effects on lung function, duration of mechanical ventilation, and mortality. Although our sample was representative of a population with sepsis-induced established ARDS resulting in MODS, this pilot study suggests that demonstrating an effect of inhNO on ARDS mortality in this population would be difficult to achieve even with a large multicenter trial. In these conditions, we suggest that further studies should focus on early treatment (when the condition is more likely to be reversible) and on the use of inhNO on selected populations such as patients with ARDS induced by direct lung injury or those responsive to the therapy whatever the etiology. These populations should benefit the most from the use of inhNO with a putative effect not only on gas exchange but also on mortality.