INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Despite recent improvements in intensive care medicine, the case-fatality rate for acute lung injury/acute respiratory distress syndrome (ALI/ARDS) remains high, at approximately 40% (1). Many patients who survive experience serious long-term health consequences, such as decreased pulmonary function and impaired health status (6). A recent National Heart, Lung, and Blood Institute ARDS Network randomized controlled trial demonstrated that a low tidal volume (VT) mechanical ventilation strategy (6 ml/kg) reduced mortality by 22% compared with traditional mechanical ventilation (12 ml/ kg) in patients with ALI/ARDS (11). However, this study did not report the relative efficacy of low VT ventilation among patients with different clinical risk factors for ALI/ARDS.

The predisposing clinical risk factor may influence the pathogenesis and clinical outcome of ALI/ARDS. Several studies implicate neutrophils as a critical mediator of lung injury from aspiration (12, 13), whereas their role in other conditions, such as pneumonia, is less certain (14). The clinical risk factor also appears to affect respiratory mechanics. Patients with ALI/ARDS resulting from pneumonia have lower lung compliance than those with extrapulmonary causes of lung injury (17). Beyond these physiologic differences, mortality differs substantially among patients with different predisposing conditions for ALI/ARDS (18). In particular, sepsis (2, 18, 22) and aspiration pneumonitis (18, 24) have been associated with the highest mortality, whereas patients with lung injury resulting from major trauma have a lower risk of death (2, 25).

Because ALI/ARDS appears to be a heterogeneous clinical condition, the low VT ventilation strategy might not equally benefit all patient subgroups. For example, the low VT strategy could have different effects in patients with pneumonia, who have lower lung compliance, than in those with other predisposing disorders. Using data from patients enrolled in ARDS Network randomized controlled trials, we retrospectively examined whether the efficacy of low VT mechanical ventilation varied by clinical risk factor for ALI/ARDS.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In the present study, we used data from 902 patients participating in ARDS Network multicenter randomized controlled trials. The protocol was approved by the institutional review board at each hospital. The results of the traditional versus lower VT ventilation trial have been reported (11). Of the 902 patients in the present study, the first 861 subjects participated in a randomized trial of low VT versus traditional VT ventilation. Using a factorial design, this trial was conducted simultaneously with two other clinical trials evaluating ketoconazole (234 patients) and lisofylline (194 patients). After the ventilator trial results became available, an additional 41 lisofylline trial subjects received 6 ml/kg in nonrandomized fashion. Of the 902 patients, 433 were randomized to ventilator group and received no ketoconazole or lisofylline.

Detailed study methods are described in the online data supplement. Briefly, intubated, mechanically ventilated patients were eligible if they met criteria for ALI or ARDS and were enrolled within 36 h. Participating subjects were randomly assigned to either the traditional (12 ml/kg) or lower VT (6 ml/kg) study group (except where noted previously).

For each patient, the clinical coordinator and physician investigator assessed the predominant clinical risk factor for ALI/ARDS within 36 h of onset. The clinical risk factor was ascertained prospectively, before randomization to ventilator treatment group. Based on careful review of the clinical and laboratory data, the predominant clinical risk factor was classified as pneumonia, sepsis, aspiration pneumonitis, trauma, or other (including drug overdose, multiple transfusion, and cardiopulmonary bypass). Risk factor classification was based on clinical judgment, rather than on strictly specified criteria. This method is likely to mirror that used in clinical practice.

Patients were followed to Day 180 or until discharge home with unassisted breathing. In the present study, the primary outcome was mortality before discharge home with unassisted breathing. Patients alive in health care facilities at 180 d were considered to have survived. Secondary study outcomes included the proportion of patients developing new nonpulmonary organ failures by Day 28, as previously defined (26). Other secondary outcomes included the proportion of patients achieving unassisted breathing by Day 28. Furthermore, we analyzed ventilator-free days, which was defined as the number of days of unassisted breathing from Day 1 to 28 if unassisted breathing continued  48 consecutive hours (11).

We used logistic regression analysis to study the primary study outcome (mortality) and secondary dichotomous outcomes (proportion achieving unassisted breathing by Day 28 and new-onset nonpulmonary organ failure). To examine whether the efficacy of low VT ventilation varied by clinical risk factor for ALI/ARDS, we tested the statistical interaction between treatment group (6 ml/kg versus 12 ml/kg VT ventilation) and clinical risk factor. We used the likelihood ratio test to compare a logistic regression model including treatment-risk factor interaction terms with a nested model including only the main effects for treatment group and clinical risk factor (i.e., no interaction terms). A statistically significant interaction term would indicate that the efficacy of low VT ventilation differed among clinical risk factor subgroups. For ventilator-free days, we used linear regression analysis in a similar fashion.

When more than one clinical risk factor for ALI/ARDS is present, the predominant predisposing condition may not always be unequivocally determined. To address this limitation of clinical classification, we reevaluated the relative efficacy of low VT ventilation after reclassifying the clinical risk factor for ALI/ARDS as a simple dichotomous category: direct pulmonary predisposing conditions (i.e., pneumonia, aspiration) and nonpulmonary conditions (i.e., sepsis, trauma, other). Furthermore, we addressed the potential overlap between pneumonia and sepsis by reclassifying the clinical risk factors as infection-related (i.e., sepsis, pneumonia) or non-infection-related (i.e., trauma, aspiration, other).

Although we analyzed data from randomized trials, a few later patients (n = 41) were nonrandomly assigned the low VT ventilation strategy. Moreover, the distribution of illness severity and other confounding factors may not be equalized between the 6 ml/kg and 12 ml/kg groups because of the smaller sample size in each clinical risk factor stratum (27). To control for these potential imbalances, we performed multivariate analysis that included baseline demographic and clinical factors that may be related to the outcome of ALI/ARDS (2, 19, 24, 28).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Baseline Characteristics

Demographic characteristics differed among the five groups defined by clinical risk factor for ALI/ARDS (Table 1). Subjects with aspiration were the oldest (mean age 55 yr), whereas persons with trauma were the youngest (mean age 44 yr). There were also significant differences by sex and race (Table 1).

There were also observed notable differences in baseline clinical variables (before ventilator group assignment) among the clinical risk groups (Table 1). Patients with trauma had the highest systolic, diastolic, and mean blood pressure, whereas subjects with sepsis and pneumonia had the lowest blood pressures. Patients with trauma had the lowest baseline prevalence of vasopressor use (17%), whereas those with sepsis had the highest prevalence (50%). Consistent with these hemodynamic findings, patients with sepsis had the highest Acute Physiology and Chronic Health Evaluation (APACHE) III scores (mean 92 points), whereas those with trauma had the lowest scores (mean 61 points).

There were differences among the clinical risk groups in every baseline respiratory variable examined except for Pa_O2 and plateau pressure (Table 1). The ratio of arterial oxygen pressure to fraction of inspired oxygen (Pa_O2/FI_O2) was highest in patients with trauma (mean 173) and lowest in those with pneumonia (mean 133).

Mortality and Clinical Risk Factor Group

The clinical risk factor for development of ALI/ARDS was associated with the case-fatality rate (i.e., cumulative incidence of death among patients with ALI/ARDS). The risk of death was highest in patients with sepsis (43%); intermediate in subjects with pneumonia (36%), aspiration (37%), and other risk factors (35%); and lowest in persons with trauma (11%) (p < 0.0001). The mortality rate for sepsis (p = 0.052) and trauma (p < 0.0001) appeared different from the other clinical risk categories.

Compared with trauma patients, subjects with sepsis (odds ratio [OR] 5.8; 95% confidence interval [CI] 2.9 to 11.4), pneumonia (OR 4.3; 2.2 to 8.3), aspiration (OR 4.4; 2.1 to 9.0), and other factors (OR 4.0; 1.9 to 8.4) had a greater risk of mortality, controlling for ventilator treatment group. To control for potential differences in acute illness severity among the clinical risk groups, we included ventilator group, age, sex, race, Pa_O2/FI_O2 ratio, APACHE III score, baseline nonpulmonary organ failure, and vasopressor use as covariates. The risk of death remained elevated in subjects with sepsis (OR 2.4; 95% CI 1.1 to 5.3), pneumonia (OR 2.7; 95% CI 1.2 to 5.8), aspiration (OR 2.2; 95% CI 1.0 to 5.1), and other factors (OR 2.6; 95% CI 1.1 to 6.1).

Given the differences in mortality among clinical risk groups, is there evidence that the efficacy of low VT ventilation varied by risk factor for ALI/ARDS? There was no statistical interaction between ventilator treatment strategy and clinical risk group (p = 0.76), providing no evidence that treatment efficacy differed among the five risk groups (Table 2). After controlling for the other covariates, there was still no evidence that efficacy varied among the clinical risk groups (p = 0.91).

To further examine mortality, we compared subjects with direct pulmonary risk factors (pneumonia or aspiration, n = 454) and those with nonpulmonary risk factors for ALI/ARDS (n = 448) (Table 3). The case-fatality rate was similar among persons with pulmonary (36%) and nonpulmonary risk factors (34%) (p = 0.57; OR 1.1; 95% CI 0.8 to 1.4). There was no statistical interaction between ventilator treatment strategy and having pulmonary or nonpulmonary risk factors, after controlling for covariates (p = 0.61). When the clinical risk factors were reclassified as infection-related (sepsis, pneumonia) or non-infection-related conditions (trauma, aspiration, other), there was also no statistical interaction between ventilator treatment and risk factor in multivariate analysis (p = 0.52).

Liberation from Mechanical Ventilation and Clinical Risk Factor Group

The proportion of patients achieving unassisted breathing by Study Day 28 varied by clinical risk group (p = 0.002) (Table 4). The probability of unassisted breathing was highest in trauma patients (76%) and lowest in persons with sepsis (51%). There was no statistical interaction between ventilator treatment group and clinical risk factor, providing no evidence that treatment efficacy varied among the groups (p = 0.59). Controlling for covariates did not appreciably influence assessment of this interaction (p = 0.61).

The mean ventilator-free days also differed by clinical risk group, with trauma patients experiencing the greatest (14.0 d), aspiration (13.1 d) and other factors (12.3 d) having intermediate values, and pneumonia (10.9 d) and sepsis (9.6 d) experiencing the fewest (p = 0.004). Although the 6 ml/kg ventilation strategy increased the number of ventilator-free days overall (mean increase 2.0 d; 95% CI 0.6 to 3.4 d), there was no interaction between ventilator group and clinical risk factor for ALI/ARDS (p = 0.58).

After reclassifying the predisposing condition for ALI/ ARDS as pulmonary or nonpulmonary causes, there was also no statistical interaction between ventilator treatment strategy and clinical risk factor group for proportion achieving unassisted breathing (p = 0.22) or ventilator-free days in multivariate analyses (p = 0.59). Similarly, reclassifying the clinical risk groups as infection-related or non-infection-related did not reveal any statistical evidence of interaction (p = 0.91 and p = 0.57, respectively).

Nonpulmonary Organ Failure and Clinical Risk Factor Group

The cumulative incidence of nonpulmonary organ failure by Day 28 was highest in patients with sepsis (67%) and lowest in those with trauma (55%) (Table 5). These differences, however, were not statistically significant (p = 0.36). We observed no statistical interaction between ventilator treatment strategy and clinical risk group (p = 0.44), even after controlling for covariates (p = 0.35). Although patients treated with 6 ml/kg VT experienced a lower risk of developing nonpulmonary organ failure (OR 0.70; 95% CI 0.5 to 0.9), there was no evidence that the efficacy of low VT ventilation varied among the clinical risk groups. Reclassifying the clinical risk factor groups as pulmonary/nonpulmonary or infection-related/non-infection-related did not reveal any statistical evidence of interaction (p  0.50 in all cases).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The clinical risk factor for ALI/ARDS was associated with substantial variation in acute illness severity and mortality. Patients with sepsis had the greatest risk of death, whereas those with trauma had the lowest risk. Despite these differences in disease severity, we found no evidence that the efficacy of the low VT ventilation strategy differed among clinical risk factor subgroups. Based on these results, the low VT strategy should be broadly applied to patients with ALI/ARDS.

The consistency of these results supports the beneficial effect of low VT ventilation in patients with diverse clinical risk factors for ALI/ARDS. There was no evidence that the efficacy of this ventilation strategy varied by clinical risk group for mortality, proportion of patients achieving unassisted breathing, ventilator-free days, or incidence of nonpulmonary organ failure. Given the range of outcomes examined, a clinically significant difference in treatment efficacy among subgroups would be unlikely.

Although there was no statistical evidence that the efficacy of low VT ventilation varied by clinical risk factor for ALI/ ARDS, simple inspection of the data might suggest no clinical benefit among patients with aspiration. Based on the lack of observed statistical interaction, chance variation may explain this apparent lack of efficacy in aspiration patients. Indeed, a rat model of acid-induced lung injury suggests that low VT ventilation improves mortality (33). Alternatively, there could be important clinical differences in aspiration-related ALI/ ARDS that attenuate the efficacy of low VT ventilation in this group. Further research is required to assess this specific patient subgroup. Because there was no statistical evidence of differential efficacy and no suggestion of harmful effect, we currently recommend treating aspiration-related ALI/ARDS with the low VT ventilation strategy until additional studies are available.

Some investigators have proposed that ALI/ARDS directly related to pulmonary disease clinically differs from that due to nonpulmonary disease (17). For instance, patients with ARDS resulting from pneumonia have lower lung compliance than those with extrapulmonary origins (17). Furthermore, direct lung injury has been associated with a greater mortality (21). In contrast, we found no difference in case-fatality rates among patients with pulmonary versus nonpulmonary risk factors for ALI/ARDS. There was also no evidence that the efficacy of low VT ventilation differed for pulmonary or extrapulmonary clinical risk groups.

The ARDS Network clinical trials were not specifically designed to evaluate the relative efficacy of the low VT ventilation strategy among clinical risk factor subgroups. Although the clinical risk factors for ALI/ARDS were defined before randomization, the present analysis was conceived after trial completion. As a result, the analysis of smaller subgroups may not fully preserve the benefits of randomization (27). As expected, the distribution of baseline demographic and clinical variables was not uniform among clinical risk factor subgroups. After statistically controlling for markers of illness severity, including APACHE III score, we still observed no evidence of differential efficacy among patient subgroups. We cannot, however, completely exclude the influence of residual confounding.

The statistical power to detect differences in efficacy among clinical risk groups poses another potential study limitation. The overall clinical trial was statistically powered to detect an absolute mortality reduction of 10%, providing lower statistical power for subgroup analysis. For example, the power to detect statistical evidence of differential treatment efficacy if low VT ventilation had no effect in the aspiration subgroup and the same average effect in all other patients was 30%. As a consequence, the lack of statistically significant differences in efficacy among clinical risk groups could represent a false-negative result (type II error). However, knowledge of statistical power, which is most useful in planning sample size before a study is conducted, may have limited utility for the clinician who is deciding whether to implement the low VT strategy in a patient with a specific clinical risk factor for ALI/ARDS. The more clinically relevant question is: given the results of our analysis, what is the probability that the low VT strategy reduces mortality in each risk factor subgroup? Based on Bayesian subset analysis (34), the probability that low VT reduces mortality in each of the five subgroups is greater than 75%. Overall, the evidence is most consistent with benefit in each subgroup. Until more is known about these subgroups, we believe that the most conservative approach is to treat all ALI/ ARDS patients with the low VT strategy.

In some cases, the classification of causes of ALI/ARDS may be difficult. Some patients have more than one potential cause, such as pneumonia that results in sepsis. Individual study coordinators and physician investigators may also apply clinical criteria differently. Although we have no formal assessment of interrater reliability, our method of classifying the clinical risk factor for ALI/ARDS is likely similar to that used in clinical practice. Furthermore, reclassifying the clinical risk factors as simpler dichotomous categoriespulmonary/nonpulmonary and infection-related/non-infection-relateddid not appreciably affect the results. The observed higher mortality in patients with sepsis is also similar to previous studies, suggesting that our classification of causes is comparable (19- 22). As a consequence, our results support the application of low VT ventilation to patients with ALI/ARDS associated with different clinical risk factors.

The introduction of any therapeutic intervention based on randomized controlled trial results depends on clinicians accepting that the study results apply to their patients. Not only did this large trial include patients with diverse clinical risk factors for ALI/ARDS, we found no difference in treatment efficacy among clinical subgroups. At present, physicians should broadly apply the low VT strategy to patients with ALI/ARDS.