INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Concern has recently been raised on the use of mechanical ventilation in acute respiratory distress syndrome (ARDS) because of its potential for generating, perpetuating, or worsening lesions of the alveolar-capillary membrane (1, 2). A number of animal studies have demonstrated that the use of large tidal volumes and pressures could produce considerable deterioration in lung function (2). A major determinant of this ventilator-induced lung injury is the level of end-inspiratory lung volume (2). Since absolute lung volume cannot be measured in clinical practice, it has been recommended to limit peak alveolar pressure, estimated at the bedside from measurement of end-inspiratory plateau pressure (1). Results of nonrandomized studies and of randomized studies with small numbers of patients have suggested that limitation of end- inspiratory pressure, a strategy also referred to as permissive hypercapnia (1, 9), could improve outcome in mechanically ventilated patients with ARDS (9).

Based on convincing animal data, it has already been recommended by a consensus conference that peak alveolar pressure should be kept below 35 cm H₂O, or that maximal transalveolar pressure should not exceed 30-35 cm H₂O. Although reasoning on transpulmonary pressure is sound, this measurement is not available in clinical practice, and peak plateau pressure has been regarded as the easiest measurement to target. An end-inspiratory plateau pressure of 35 cm H₂O has been frequently proposed as the upper tolerable limit in patients with ARDS (1).This approach has already reached current practice. Several animal studies, however, have shown that lung injury can occur with airway pressures at or even below 30 cm H₂O (8, 12), and that pre-existing lung injury may have a synergistic action with mechanical ventilation (13). Besides, careful examination of the pressure-volume relationship of the respiratory system in patients with ARDS has suggested that lung overdistension may be occurring at much lower levels of plateau pressure (14). Based on published values of the upper inflection point of the pressure-volume curvethought to signal overdistentiontargeting plateau pressure at 25 cm H₂O as the upper limit for the titration of tidal volume during mechanical ventilation of patients with ARDS has been suggested (14). Human data clearly supporting the permissive hypercapnia approach are lacking, however, whereas the hypercapnia and acidosis that result from reduction in minute ventilation as well as the possible risk of alveolar collapse that could result from such settings could be hazardous (17). Accordingly, we undertook a prospective multicenter, randomized study to compare a conventional approach to mechanical ventilation, reproducing the actual routine practice in many centers at the time of the study with a strategy aimed at further limiting the plateau pressure, with the goal of reducing mortality in patients presenting with ARDS and no other organ failure at entry.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The study was conducted between January 1994 and September 1996 in the intensive care units of 25 hospitals in seven countries (listed in the Appendix ). The study protocol was approved by each institutional review board. Before enrollment a signed consent form was obtained from the patient, whenever possible, or from the patient's next of kin or legal representative.

Patients above 17 and below 76 y of age were included if they presented with diffuse bilateral infiltrates on the chest X-ray, had arterial hypoxemia requiring mechanical ventilation and the use of an inspired fraction of oxygen of 0.50 or greater for at least 24 h, and a Lung Injury Score above 2.5 for less than 72 h (18). This score, used to define ARDS, was computed from the number of quadrants of the chest X-ray displaying opacities, the degree of hypoxemia, the level of positive end-expiratory pressure (PEEP), and the respiratory system compliance; the average score was used to define ARDS as previously described (18).

Patients were not included in the study if one of the following was present: a history of left heart failure, or evidence of cardiogenic edema or elevated pulmonary capillary wedge pressure ( 18 mm Hg); presence of severe organ failure other than the lung (cardiovascular, renal, hematologic, or neurologic), as defined by Knaus and colleagues (19), or need for high levels of vasopressive agents (epinephrine > 1 mg/h or norepinephrine > 2 mg/h); presence of a pre-existing chronic disease, using the definitions of the Acute Physiology and Chronic Health Evaluation (APACHE) II (19) (obstructive lung disease, liver failure, or renal failure), a moribund state, or presence of AIDS; presence of morbid obesity, major chest wall abnormalities (kyphoscoliosis or open or flail chest), or a chest tube in place with persistent air leak; bone marrow transplant; or intracranial hypertension or head injury with obtundation, defined as a decrease in score by at least two points using the Glasgow Coma Scale.

Randomization and Treatment Groups

To ensure that causes of ARDS that are likely to influence mortality would be equally distributed between the two arms of the study, patients were stratified into three groups before randomization: (1) multiple trauma; (2) immunosuppressive therapy (organ transplant, malignancies, or steroid-dependence); and (3) others. Randomization was then performed by center using the sealed-envelope method to allocate the patient to one of the two arms.

Titration of positive end-expiratory pressure. Since the aim of the study was to compare two different tidal volume settings with little or no difference in other parameters, the selection of PEEP was made before randomization. A "PEEP trial" was performed using increments of 5 cm H₂O (from 0 to 15) during pure oxygen breathing to determine the optimal level of PEEP. This trial sought the greatest improvement in oxygenation or the first level allowing the Pa_O2/FI_O2 ratio to be above 200 mm Hg without worsening hemodynamics (defined as a 20% drop in cardiac output or blood pressure or a 20% increase in arteriovenous oxygen content difference). Airway occlusion was performed at end-expiration to measure end-expiratory alveolar pressure, also termed total PEEP (20). Total PEEP was used throughout the study.

Standard treatment group. Mechanical ventilation was delivered using standard volume-targeted (assist-control) ventilation, with a tidal volume of 10 ml/kg of body weight or above (up to 15), and a respiratory rate adjusted to maintain the partial pressure of arterial carbon dioxide (Pa_CO2) between 38 and 42 mm Hg; there was no requirement for peak flow setting, but the ratio of inspiration to expiration was never higher than 1. As a safety feature, because of concern for the risk of barotrauma (21, 22), tidal volume was not increased when peak airway pressure reached 60 cm H₂O, irrespective of the Pa_CO2.

Pressure limitation group. Also using assist-control ventilation, tidal volume was primarily titrated to maintain the end-inspiratory plateau pressure at or below 25 cm H₂O; plateau pressure was measured after a 2-s pause or longer when patients were relaxed, not coughing, or moving. Tidal volume was maintained at less than 10 ml/ kg but not lower than 6 ml/kg or 300 ml, irrespective of the plateau pressure, as a safety limit to avoid excessive hypoventilation. Tidal volume could be increased up to a plateau pressure of 30 cm H₂O if inspired oxygen fraction was at 0.9 or above, if reduced chest wall compliance was suspected, or if major acidosis was present (pH below 7.05). There was no recommendation for respiratory rate or peak flow settings. If pH decreased below 7.05, a careful titration using sodium bicarbonate was recommended. When metabolic acidosis was associated, a rapid correction was recommended, e.g., using dialysis for renal failure.

In both groups, body weight was defined as actual body weight minus the estimated weight gain due to water and salt retention, which is frequently present in septic patients.

Other treatments. Sedation and paralysis could be used as required. The use of inhaled nitric oxide was allowed. The minimal FI_O2 was used to maintain arterial oxygen saturation at or above 92% when FI_O2 was at or above 0.8. When a FI_O2 below 0.8 was used, Sa_O2 could be maintained up to 95% without decreasing FI_O2. With FI_O2 at or below 0.6, Sa_O2 could be even higher. Direct arterial blood measurements of oxygen saturation were required for this adjustment (not pulse oxymetry). The level of ventilator support was decreased as the patient's status improved; the same upper limit for plateau pressure was used in the pressure limited group, even if other modes were used, such as pressure support ventilation.

Data Collection

In addition to recordings of arterial blood gases (recorded once a day) and ventilator settings, APACHE II (19) and Simplified Acute Physiology Score (SAPS) II (23) were calculated both on admission to the intensive care unit and on inclusion in the study; the number of organ systems failing were calculated on Days 1, 3, 7, and 14 (19).

Several meetings of the investigators were organized with the study coordinator (L.B.) to discuss in depth all aspects of the protocol. Every center had at least one main investigator (and usually one or two other investigators), who checked the protocol and the correct application of the procedures every day (all investigators among countries had already conducted or participated in clinical trials). Every inclusion was reported within 48 h to the study methodologist (F.R.T.). The case report form for recording physiologic data was filled out daily. For calculated indexes such as SAPS or Organ System Failure score (OSF), the range of values was recorded for SAPS or the exact value for OSF. All case report forms were carefully reviewed (L.B., F.R.T., C.B.B.) to check for inconsistency, errors, or violation of the protocol. All problems, including errors, missing data, or inconsistency, were solved by conversations over the phone.

Statistics

End points. The primary outcome variable was a 60-d mortality. Secondary variables included incidence of pneumothorax requiring placement of a chest tube, incidence of secondary organ system failures, length of mechanical ventilation, and length of stay in the intensive care unit. We also calculated the number of ventilator-free days up to Day 60 in both groups, as recently proposed (24).

Methods. A total number of 240 patients was planned to demonstrate a reduction in mortality from 50 to 30%, with an alpha risk of error of 5% and a beta risk of error of 10% using a bilateral approach. A sequential analysis was used and interim analyses were performed every 20 patients using the triangular test, which allows a study to be stopped at an early stage when a boundary has been crossed after computation of the mortality rate (25). Crossing the upper boundary signifies that the new strategy works better, whereas crossing the lower boundary signifies that a benefit in mortality will not be demonstrated, even after including the total number of planned patients (25). The study was stopped prematurely after an interim analysis performed on the first 100 patients. Kaplan-Meier survival curves were compared using the log-rank test. A chi-square test or Fisher exact test was used when necessary for two-by-two comparisons of categorical data. Nonparametric tests (Mann-Whitney test and Kruskal-Wallis ANOVA) were used for analysis of quantitative data. All analyses were performed using BMDP software (1D, 3D, 4F, 3S, 1L) (26). A p value of less than 0.05 was considered to indicate statistical significance.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

At the time of study termination, 116 patients had been enrolled in the study (58 patients in each group). There was no significant difference in the characteristics of patients on admission to the intensive care unit and on inclusion in the study (Table 1). Additional treatments were also similar: sedation was administered for 11 ± 9 d in both groups, and paralysis was required in 43 patients (74%) in the pressure limitation group versus 34 (59%) in the standard group (p = 0.12); nitric oxide was given to 16 (28%) and 14 (24%) patients, respectively (p = 0.67). The mean delay between the beginning of ARDS and inclusion was 1.2 ± 0.8 d in the pressure limitation group and 1.1 ± 0.9 d in the standard group. In the first group 26% of the patients were enrolled after more than 24 h and 33% in the second group. None of these differences was significant.

Table 2 lists the mean ventilatory settings and arterial blood gas values in the two groups on Days 1, 2, 7, and 14. As expected by study design, ventilatory settings differed significantly between the two groups, except for PEEP, which was set identically in the two groups. Figures 1 and 2 show the individual distribution of tidal volumes and plateau pressures in the two groups, indicating some overlapping due to the safety limits imposed in each arm. The case report forms were reviewed to check whether the patients were ventilated following the rules of the protocol. Specifically, the question was: in the case of pressures and volumes that were not in the range dictated by the arm of the study, was it due to a safety limit imposed by the protocol? The answer was positive in all cases, except for two patients in the standard group, where the peak pressures were at 55 and 58 cm H₂O on the day the volume was limited, which was found to be acceptable. In the new treatment group, 97% of the patients were ventilated with tidal volumes below 10 ml/kg, and 90% had a plateau pressure not greater than 30 cm H₂O; 59% of patients had a plateau pressure not greater than 25 cm H₂O in the experimental group, while it was the case for only 19% of patients in the standard group. Figure 3 shows the evolution of oxygenation in the two groups, with no significant difference over time. Figure 4 shows the evolution of PEEP in the two groups, which was progressively reduced in a similar manner in the two groups.

View larger version (12K): [in this window] [in a new window]   Figure 1.   Individual values of tidal volume at Day 1 in the two groups. Patients alive at Day 60 are indicated by open circles; patients deceased are indicated by closed circles. The horizontal bars indicate the mean values. Ninty-seven percent of the patients in the pressure limitation group were ventilated with tidal volumes below 10 ml/kg, whereas 83% of the patients in the standard group had volumes at or above this threshold.

View larger version (12K): [in this window] [in a new window]   Figure 2.   Individual values of plateau pressures at Day 1 in the two groups. The horizontal bars indicate the mean values.

View larger version (15K): [in this window] [in a new window]   Figure 3.   Mean (± standard deviation) values of PaO2/FIO2 over the first 14 d of the study in the two groups. The number of patients still alive and mechanically ventilated at each day is indicated for both groups.

View larger version (12K): [in this window] [in a new window]   Figure 4.   Mean (± standard error of the mean) values of the PEEP level over the first 14 d of the study in the two groups. The same numbers of patients indicated in Figure 3 were used for each day.

The study was stopped after interim analysis of the first 100 patients included, using the sequential approach, because it was calculated that pressure limitation would not achieve a beneficial effect within the planned frame of the study. Figure 5 shows the Kaplan-Meier survival curve in the two groups. No significant difference between the two groups could be observed using the log-rank test (p = 0.39). Mortality at Day 60 was not different between the two groups (46.6% in the new treatment group versus 37.9%, p = 0.38). Figure 6 shows the survival curves in patients depending on whether they developed multiple organ failure, defined as the failure of at least one organ other than the lung. Again, these subgroups displayed no difference in mortality between the two arms.

View larger version (12K): [in this window] [in a new window]   Figure 5.   Survival curves in the two groups using the Kaplan-Meier analysis.

View larger version (16K): [in this window] [in a new window]   Figure 6.   Survival curves in the two groups of patients stratified by the presence (MOF+) or absence (MOF) of secondary multiple organ failure.

Table 3 indicates the results of secondary outcome variables between the two groups. No significant differences were observed.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In mechanically ventilated patients with ARDS, we found that a deliberate reduction in tidal volume to achieve an end-inspiratory plateau pressure around 25 cm H₂O did not appear to have a significant impact on morbidity and mortality, as compared with a more conventional ventilatory strategy where normocapnia was achieved with tidal volumes in the range of 10 to 11 ml/kg and plateau pressures already below 35 cm H₂O.

The underlying hypothesis of this study was that administering mechanical ventilation with the classic objective of normalizing alveolar ventilation may result in excess mortality in ARDS because of ventilation-induced alveolar trauma. In addition to evidence from animal studies, this hypothesis has been supported indirectly by the presence of cysts, bullae, or emphysematous-like lesions at autopsy of such patients (21, 22, 28, 29). This hypothesis is not supported by the data of the present study, perhaps because very high volumes and pressures were not employed in the control group. Although our two groups differed significantly in terms of tidal volume, the mean plateau pressure in the control group actually remained below 35 cm H₂O (Table 2), the arbitrary threshold espoused by a consensus conference as the upper limit above which there could be a high likelihood of ventilator-induced injury (1). While tidal volumes of 15 ml/kg or higher were frequently employed several years ago (31), this approach was no longer considered acceptable at the time we designed our study (1, 31). It is likely that the relationship between mechanical ventilation-induced alveolar damage and the volume or pressure used is not linear, as suggested by experimental studies (13). It is possible that both patient groups in the present study were ventilated in a volume-pressure range below that associated with hyperinflation and alveolar trauma. Indeed, the low incidence of pneumothorax in both arms of the study supports this viewpoint. For instance, 95% of the patients in the pressure limitation group were ventilated with a plateau pressure below 35 cm H₂O compared with more than 72% of the patients in the standard group. As such, the present study helps to define a safe zone for mechanical ventilation, avoiding undue hypercapnia and respiratory acidosis, on the one hand, and the risk for ventilator-induced lung injury on the other hand.

Many aspects of the study limit its interpretation. This study was not powered to detect small differences in outcome between the two groups, such as 5% differences in mortality or in ventilator-free days. In addition, the absence of precise protocol for weaning from mechanical ventilation makes it difficult to compare small differences in the duration of mechanical ventilation. The reduction of ventilatory assistance looked very similar in the two groups, however, as illustrated by the progressive reduction in PEEP level (Figure 4). This suggests that no major differences existed in terms of reducing the amount of support in the two arms. In addition, no trend existed in any of the outcome indexes in favor of the experimental arm, suggesting that the number of patients needed to find a difference, if any, would be extremely large.

Our results contrast with those of Amato and coworkers (32, 33) who compared two approaches to mechanical ventilation of patients with ARDS and found a new approach limiting volumes and pressures to be beneficial. In their first report both oxygenation and lung mechanics were found to be improved using this new approach (32), while their final result in a group of 48 patients indicated improved survival (33). It is noteworthy, however, that their two approaches differed in many aspects, including PEEP level, mode of ventilation, and end-inspiratory plateau pressure, making it difficult to assess the respective importance of each factor. It is also interesting to note that they used tidal volumes in the range of 12-13 ml/ kg in the control group. In this group, plateau pressures were close to 40 cm H₂O after a few days, and this was associated with a 42% rate of pneumothorax, much higher than in the two arms of our study. Whether the high plateau pressures they have used explain part of the 70% mortality rate at 30 d in their control group is difficult to ascertain, but this supports the hypothesis that a volume or pressure threshold higher than that targeted in the control group of our study needs to be reached to observe an excess in mortality.

The pressure-volume curve of the respiratory system provides information on the volume of aerated lung that can be ventilated (34). This test can also provide insights into the response of the respiratory system to increasing tidal volumes (14, 35, 36). In a group of 25 mechanically ventilated patients with ARDS, we previously found that the mean level of pressure at which the curve became nonlinear and respiratory system compliance decreased (i.e., the upper inflection point of the pressure-volume curve), was as low as 25 cm H₂O on average (14). We thus hypothesized that ventilating all patients with an end-inspiratory plateau pressure above 25 cm H₂O would put more than 80% of patients at risk of hyperinflation. Moreover, in a subsequent study using computed tomography scanning we found that ventilating the lungs of patients with ARDS above this upper inflection point primarily resulted in further hyperinflation without substantial additional recruitment (16). These data represented a link between data accumulated in experimental animals (2) and the notion that end inspiratory hyperinflation could be a frequent occurrence in patients being ventilated for ARDS, and prompted the present study. The negative results of the present study suggest that such an approach was probably an oversimplification. We did not use the curves here, however, and the lack of benefit of limiting tidal volume in our study may be related to the lack of individual titration of ventilation. Because obtaining pressure-volume curves is technically demanding and interpretation of the data can be complex, we decided against having physicians perform a procedure of which they had limited experience, and instead selected an upper limit of plateau pressure that would minimize the likelihood of ventilatory excursions reaching the upper inflection point in most patients. The range of pressure for individual patients can be highly variable, however, and the benefits offered to some patients may be largely counterbalanced by unnecessary deleterious effects in others.

Side effects of permissive hypercapnia certainly exist, though the precise limits of clinical tolerance have not been well defined (17). Arterial oxygen desaturation (37), alveolar instability (38), pulmonary hypertension (39), and modification of regional perfusions (17) are among the possible drawbacks of this strategy. We did not find any difference in morbidity between the two groups, but one may hypothesize again that the benefits obtained by limiting the pressure in this study were obscured by hidden side effects resulting from hypoventilation. Insufficient alveolar recruitment due to excessive reduction in tidal volume is a particular concern because animal studies indicate that it may in itself cause alveolar injury (40, 41). Because alveolar recruitment occurs during tidal inflation (42), reduction of tidal volume may prove beneficial when it prevents hyperinflation, but it could be quite harmful if it is unable to recruit previously collapsed or compressed alveoli. The level of positive end-expiratory pressure may be especially important in this situation, as shown by experimental (38, 39) and human data (43). In our study no difference existed in terms of PEEP between the two groups. Using much higher levels of PEEP in the new treatment group than recommended with the standard approach (1, 31) could possibly prevent the untoward effects of low tidal ventilation on alveolar de-recruitment.

We found that a reduction in delivered tidal volume to limit plateau pressure to approximately 25 cm H₂O failed to favorably influence any outcome variable examined in this group of patients with ARDS when compared with a group where normocapnia was achieved with plateau pressures already below 35 cm H₂O. In both groups the incidence of clinical barotrauma was low, indirectly suggesting that major overdistension was not a frequent event in any of the two groups.