Higher mini-BAL total protein concentration in early ARDS predicts faster resolution of lung injury measured by more ventilator-free days

Abstract

The protein concentration of alveolar edema fluid in acute respiratory distress syndrome (ARDS) is dynamic. It reflects alveolar flooding during acute injury, as well as fluid and protein clearance over time. We hypothesized that among ARDS patients treated with low tidal volume ventilation, higher concentrations of protein in mini-bronchoalveolar lavage (mBAL) samples would predict slower resolution of lung injury and worse clinical outcomes. Total protein and IgM concentrations in day 0 mBAL samples from 79 subjects enrolled in the aerosolized albuterol (ALTA) ARDS Network Albuterol Trial were measured by colorimetric assay and ELISA, respectively. Linear regression models were used to test the association of mBAL proteins with clinical outcomes and measures of length of illness, including ventilator-free days (VFDs). Median mBAL total protein concentration was 1,740 μg/ml [interquartile range (IQR): 890–3,170]. Each 500 μg/ml increase in day 0 mBAL total protein was associated with an additional 0.8 VFDs [95% confidence interval (CI): 0.05–1.6, P value = 0.038]. Median mBAL IgM concentration was 410 ng/ml (IQR: 340–500). Each 50 ng/ml increase in mBAL IgM was associated with an additional 1.1 VFDs (95% CI 0.2–2.1, P value = 0.022). These associations remained significant and were not attenuated in multivariate models adjusted for age, serum protein concentration, and vasopressor use in the 24 h before enrollment. Thus, higher mBAL total protein and IgM concentrations at day 0 are associated with more VFDs in patients with ARDS and may identify patients with preserved alveolar epithelial mechanisms for net alveolar fluid clearance.

the acute respiratory distress syndrome (ARDS) is characterized by increased lung vascular permeability that results in the extravasation of protein-rich fluid into the alveolar space. More severe clinical presentations may be associated with increased vascular permeability to edema fluid and plasma proteins, as well as a loss of size selectivity of the protein gradient between plasma and alveolar compartment (13, 14, 32). Many studies have identified plasma biomarkers that predict clinical outcomes after ARDS (2, 3, 7). Fewer ARDS studies have measured biomarkers from the alveolar compartment of patients with ARDS (3, 8, 10, 17, 24). Many of these studies sampled alveolar fluid at later time points after the diagnosis of ARDS, and most were conducted when mechanically ventilated patients routinely received higher tidal volumes (8, 10, 17, 24). Notably, subjects enrolled in the ARDS Network ALVEOLI and activated protein C (APC) Trials received low tidal volume ventilation during the study period (3). Before the era of lung-protective ventilation, higher total protein concentration in bronchoalveolar lavage (BAL) from ARDS patients at day 3 and later was associated with higher mortality. Many animal models use BAL total protein as a surrogate outcome in acute lung injury (5, 21, 29, 33). Indeed, protein-rich alveolar edema fluid is a hallmark of ARDS; however, the implications of the protein concentration in alveolar fluid for clinical outcomes are likely dependent on when in the course of illness the sample is obtained, as well as the underlying severity of illness.

Alveolar fluid composition is dynamic, reflecting not only the flooding during acute injury, but also net alveolar fluid and protein clearance. Active transport of sodium across the alveolar epithelium drives fluid clearance. Protein clearance is a slower process that occurs primarily through paracellular routes in a size-dependent fashion (11, 12). Although higher protein concentration could indicate more severe lung injury, it may also reflect net removal of alveolar edema fluid and injury resolution (18). It also stands to reason that the larger proteins, such as immunoglobulin M (IgM) (970 kDa), would be cleared more slowly than small proteins, such as albumin (67 kDa), the dominant protein present in plasma and alveolar fluid (13). Small amounts of IgM may be secreted into the bronchial tree or alveolar space under normal conditions (31). This molecule is too large to pass from the plasma through tight junctions to the alveolar space under normal conditions. However, in the setting of alveolar damage, larger proteins may enter the airspaces in higher concentrations. The prognostic value of increased IgM concentrations in BAL fluid has not been studied in ARDS patients treated with a low tidal volume ventilation strategy, and only one prior study has reported findings from mBAL biomarkers in patients treated with low tidal volumes (3).

We hypothesized that higher concentrations of total protein and IgM in mini-bronchoalveolar lavage (mBAL) samples obtained early in ARDS would correlate with worse pulmonary physiological abnormalities and predict slower resolution of lung injury and worse clinical outcomes in patients treated with lung-protective ventilation. Furthermore, we hypothesized that protein concentrations in mBAL samples would correlate with previously measured plasma biomarkers in the ALTA Trial, including IL-6 and IL-8.

MATERIALS AND METHODS

ARDS Network ALTA study.

This is a nested cohort study within the larger prospective randomized multicenter trial of Aerosolized Albuterol vs. Placebo in Acute Lung Injury (ALTA) conducted by ARDS Network investigators between August 2007 and September 2008 (19). A total of 282 subjects were enrolled before the Data Safety and Monitoring Board (DSMB) stopped ALTA for futility. The institutional review board (IRB) at each site approved the study, as did the National Heart, Lung and Blood Institute DSMB. The University of California San Francisco IRB and the NIH NHLBI ARDS Network Pathogenesis Committee approved this substudy.

Mini-BAL.

The ALTA study protocol included a mini-bronchoalveolar lavage (mBAL) procedure at baseline. Study personnel performed this procedure in the Intensive Care Unit (ICU) after informed consent was obtained from the subject or their surrogate; informed consent was obtained within 48 h of meeting all three criteria for ARDS. A catheter was inserted into the subject’s endotracheal tube through a standard bronchoscopy adapter, and the inner catheter was advanced blindly into a distal airway. Normal saline (40 ml) was injected and then aspirated with gentle suction for 10 s to recover as much of the instilled fluid as possible with expected yields of 5–10 ml. Exclusion criteria for the mBAL portion of the protocol included $\mathrm{F}{\mathrm{I}}_{{\mathrm{O}}_{\mathrm{2}}}$ greater than 0.8 and peak end expiratory pressure (PEEP) >14 cm H₂O, hemodynamic instability (despite fluid resuscitation and vasopressor support), open external ventricular device or intracranial pressure greater than 15 mmHg, INR greater than 2.0 within 36 h of mBAL, platelets <50 × 103/mm³ within 36 h of mBAL. Before the mBAL portion of the study protocol was discontinued, 82 subjects underwent mBAL on the day of study enrollment (day 0), and 79 of these had detectable levels of total protein in the mBAL samples.

Biological sample collection and processing.

Blood and mBAL specimens were obtained at baseline. In brief, EDTA and citrate anticoagulated plasma, as well as mBAL lavage specimens, were centrifuged, divided, and frozen at −80°C immediately after collection and sent to the central repository for storage. All samples were transferred to the University of California, San Francisco, from the National Heart Lung and Blood Institute.

Biomarker measurements.

Biomarker measurements were made in duplicate on frozen mBAL samples. Total protein concentration was measured by a colorimetric assay (Bio-Rad, Hercules, CA), and Immunoglobulin M (IgM) was measured using a commercially available ELISA kit (Abcam, Cambridge, MA). Patients with undetectable mBAL total protein measurements were excluded on the basis of unacceptable alveolar lavage fluid sampling (n = 3). Serum protein was recorded on all ALTA patients who were also enrolled in the ARDS Network Enteral Omega-3 Fatty Acid, γ-Linolenic Acid, and Antioxidant Supplementation in Acute Lung Injury (Omega-3) Randomized Trial (n = 33). Where available, colorimetric assays on plasma samples from ALTA patients available at UCSF from prior ARDS Network substudies were used to measure total protein (n = 30). For subjects without recorded or measured serum protein concentrations [n = 16 (20%)], we used a multiple imputation approach to address missing data (15). Linear regression was used for the imputation model for serum protein. The imputation model included all independent variables used in the final multivariate analysis models (age, mBAL total protein, vasopressor), as well as the outcome variable of interest, VFDs, and variables that are potential predictors of missingness or serum protein concentration (sex, smoking status, and hemoglobin). Twenty imputations were performed. Plasma IL-6 and IL-8 were measured at baseline, as previously described (19).

Clinical outcomes, physiological parameters, and missing data.

The primary outcome of the study was ventilator-free days (VFDs), defined as the number of days to day 28 that the subject achieved unassisted breathing. Subjects who did not survive to 28 days were assigned zero VFDs. We believe that mBAL biomarkers measured early in the course of illness are unlikely to have an effect on clinical outcomes beyond 28 days and a priori chose to limit the VFD outcome to 28 days. Secondary outcomes included 28-day mortality, oxygenation index, $\mathrm{P}{\mathrm{a}}_{\mathrm{O}}{}_{{}_{\mathrm{2}}}$/$\mathrm{F}{\mathrm{I}}_{{\mathrm{O}}_{\mathrm{2}}}$ ratio, driving pressure, hospital days, and ICU days. Oxygenation index, $\mathrm{P}{\mathrm{a}}_{\mathrm{O}}{}_{{}_{\mathrm{2}}}$/$\mathrm{F}{\mathrm{I}}_{{\mathrm{O}}_{\mathrm{2}}}$ ratio, and driving pressure were calculated from the study enrollment variables obtained within the 24 h preceding mBAL. If data were missing from study enrollment variables, we used values from a separate data collection form for calculating Acute Physiology and Chronic Health Evaluation III (APACHE III) scores obtained within 24 h preceding randomization. $\mathrm{P}{\mathrm{a}}_{\mathrm{O}}{}_{{}_{\mathrm{2}}}$/$\mathrm{F}{\mathrm{I}}_{{\mathrm{O}}_{\mathrm{2}}}$ ratio was calculated by dividing the partial pressure of arterial oxygen by the fraction of inspired oxygen. Oxygenation index was calculated with the study enrollment variables, obtained at baseline, as the product of the mean airway pressure, and the fraction of inspired oxygen was divided by the partial pressure of oxygen. Driving pressure was calculated with the study enrollment variables, obtained at baseline, as the plateau pressure minus the PEEP. There was 10% missing data (n = 8) for both oxygenation index and driving pressure.

Statistical methods.

Continuous variables were expressed as means ± SD or median with interquartile range (IQR). Student’s t-test was used to test for differences of mBAL biomarkers, physiological parameters, and clinical outcomes between ALTA treatment arms. Spearman’s correlation coefficient was used to describe the association between mBAL biomarkers with each other, previously measured plasma biomarkers, and physiological parameters. Univariate and multivariate linear models were used to test the association of mBAL total protein and IgM concentrations with measures of resolution of lung injury clinical outcomes. All regression analyses except those testing associations with ventilator-free days excluded subjects who died. Multivariable logistic regression was performed by a manual step-wise backward selection approach using likelihood ratio testing with appropriate model checking and transformation of variables as needed. Initial models included adjustment for the treatment arm of the ALTA-randomized clinical trial, APACHE III score, etiology of lung injury, vasopressor use in the 24 h before enrollment, age, and serum protein concentration. Adjustment for study treatment arm, APACHE III, and etiology of lung injury was eliminated from the final models with likelihood ratio testing. All final multivariate models were adjusted for age, serum protein concentration, and vasopressor use in the 24 h before enrollment. Of note, not only is vasopressor administration a marker of severity of illness, but a study conducted before lung-protective mechanical ventilation became the standard of care showed that alveolar fluid clearance in subjects with lung injury is reduced among subjects with shock who were treated with vasopressors (34). Linearity assumptions were checked in both univariate and multivariate models. Checks included smooth transformation of mBAL biomarkers, and multivariate models were checked using component plus residual plots. Normality of residuals was checked using graphical methods. Robust standard errors or bootstrap methods were used to accommodate violations of the assumption of normality or constant variance. If both of these approaches were tested, the most conservative, or widest, confidence intervals were reported here. Bootstrap methods were used to conclude statistical significance. Each model was checked for influential points, and outliers were identified using DFBETA statistics. Excluding outliers did not substantively change the model output, and all observations were included in the reported findings. To accommodate the distribution of ventilator-free days, a dependent variable with a high zero count, we analyzed the data with both linear regression and ordinal logistic regression, a more robust model with fewer assumptions. The results of all of our model checks with both ordinal logistic regression and linear regression using transformed biomarker predictors and excluding subjects who died were similar. The transformations and alternative models used for checking our results did not substantively change the model output, and we report the findings from the untransformed predictor for a standard linear regression model. Univariate and multivariate logistic models were used to test the association of mBAL total protein and IgM concentrations with 28-day mortality. A two-sided P < 0.05 was considered significant. All analyses were performed using Stata version 13 (StataCorp, College Station, TX).

RESULTS

Baseline characteristics for the 79 subjects enrolled in the ALTA trial who underwent the mBAL protocol with detectable levels of protein in the stored specimen are shown in Table 1. Subjects had a mean age of 49 and were predominantly white (63%) and male (57%). Within this cohort, there was a mix of direct and indirect primary etiology of lung injury: pneumonia (41%), aspiration (22%), and sepsis (23%). Forty-two percent of subjects were receiving vasopressors in the 24 h before enrollment, and the median APACHE III score was 86 (±22). The median tidal volume on study enrollment was 6.2 ml/kg predicted body weight (IQR: 5.9–6.9). The 28-day mortality was 19%.

Table 1.

Characteristics of 79 ALTA subjects who underwent mBAL

Parameter	Value
Age, yr	49 ± 16
Male sex	45 (57)
Race
White	50 (63)
Black	13 (16)
Asian	4 (5)
Unknown or other	12 (15)
Latino ethnicity	6 (8)
Primary etiology of lung injury
Pneumonia	32 (41)
Sepsis	18 (23)
Aspiration	17 (22)
Trauma	6 (8)
Unknown or other	6 (8)
APACHE III score	86 (22)
Albuterol intervention arm	42 (53)
Vasopressor use	33 (42)
Serum protein, g/dl‡	5.2 (1.3)
mBAL total protein, µg/ml	1,793 (948–3221)
mBAL IgM, ng/ml	413 (335–502)
IL-6, pg/ml	259 (59–926)
IL-8, pg/ml	74 (38–168)
Mean airway pressure, cmH2O	14 (11–16)
Oxygenation Index†	7 (4–10)
$\mathrm{P}{\mathrm{a}}_{\mathrm{O}}{}_{{}_{\mathrm{2}}}$/$\mathrm{F}{\mathrm{I}}_{{\mathrm{O}}_{\mathrm{2}}}$†‡	190 (154–260)
Driving pressure†‡	13 (11–16)
Ventilator-free days*	21 (0–25)
Hospital days*	19.5 (13–28)
ICU days*	13 (9–22)
Mortality at 28 days	15 (19)

All data are presented as means ± SD, median (IQR), or n (%) as appropriate. ALTA, Aerosolized Albuterol Versus Placebo in Acute Lung Injury ARDS Network trial. APACHE, Acute Physiology and Chronic Health Evaluation; mBAL, mini-bronchoalveolar lavage; ICU, intensive care unit.

$\text{Oxygenation index}=\frac{(\text{mean airway pressure})\times (\text{fraction of inspired oxygen})}{(\text{partial pressure of arterial oxygen})}$ $\mathrm{P}{\mathrm{a}}_{\mathrm{O}}{}_{{}_{\mathrm{2}}}$/$\mathrm{F}{\mathrm{I}}_{{\mathrm{O}}_{\mathrm{2}}}$, partial pressure of arterial oxygen:fraction of the inspired oxygen. Driving pressure = plateau pressure − peak end expiratory pressure.

^*To 28 days.

^†Variable measured when subject on study protocol; see materials and methods.

^‡Missing data: Serum protein, n = 16 (20%); oxygenation index and driving pressure, n = 8 (10%).

Median day 0 mBAL total protein concentration was 1,793 μg/ml (IQR: 948–3,221). Median day 0 mBAL IgM concentration was 413 ng/ml (IQR: 335–502). The ALTA trial was a negative trial, and no significant differences in biomarker levels or clinical outcomes between treatment arms were observed. The mBAL total protein and IgM concentrations were not significantly different between the treatment arms of the ALTA trial (2,184 μg/ml vs. 1,931 μg/ml, P = 0.45 and 415 ng/ml vs. 429 ng/ml P = 0.60, respectively). VFDs to 28 days were not significantly different between the albuterol treatment arm and placebo [21.5 (0–25) vs. 20 (10–23), P value = 0.91]. In the larger ALTA cohort, the median (IQR) for VFDs in the treatment arm was 20 (0–24.5) vs. 21 (7–24) in the placebo group. With respect to VFDs, the subgroup of ALTA patients included in this study appears similar to the larger ALTA study population. Oxygenation index, $\mathrm{P}{\mathrm{a}}_{\mathrm{O}}{}_{{}_{\mathrm{2}}}$/$\mathrm{F}{\mathrm{I}}_{{\mathrm{O}}_{\mathrm{2}}}$, driving pressure, hospital days, ICU days, ventilator days, and 28-day mortality did not differ between treatment arms (data not shown).

Mini-BAL total protein and mini-BAL IgM were moderately correlated (rho 0.53, P < 0.0001) (Fig. 1). Weak correlations between mini-BAL total protein concentration oxygenation index and $\mathrm{P}{\mathrm{a}}_{\mathrm{O}}{}_{{}_{\mathrm{2}}}$/$\mathrm{F}{\mathrm{I}}_{{\mathrm{O}}_{\mathrm{2}}}$ approaching significance were observed in this nested cohort (rho 0.21, P = 0.074 and rho −0.23, P = 0.052, respectively) (Table 2). However, there was no significant correlation between mini-BAL total protein concentration and driving pressure, plasma IL-6, or plasma IL-8. There was no significant correlation between mini-BAL IgM concentration and oxygenation index, $\mathrm{P}{\mathrm{a}}_{\mathrm{O}}{}_{{}_{\mathrm{2}}}$/$\mathrm{F}{\mathrm{I}}_{{\mathrm{O}}_{\mathrm{2}}}$, driving pressure, tidal volume, plasma IL-6, or plasma IL-8.

Fig. 1.

Scatterplot of day 0 mini-bronchoalveolar lavage (mBAL) biomarkers with best-fit line and Spearman’s correlation coefficient show moderate correlation between mBAL IgM and total protein concentrations.

Table 2.

Association of day 0 mBAL total protein and IgM with physiological parameters and plasma biomarkers

	mBAL Total Protein		mBAL IgM
	Rho	P Value	Rho	P Value
Oxygenation index	0.21	0.074	−0.06	0.64
$\mathrm{P}{\mathrm{a}}_{\mathrm{O}}{}_{{}_{\mathrm{2}}}$/$\mathrm{F}{\mathrm{I}}_{{\mathrm{O}}_{\mathrm{2}}}$	−0.23	0.052	0.07	0.54
Driving pressure	−0.04	0.72	−0.14	0.25
Tidal volume*	0.13	0.37	0.12	0.28
Plasma IL-6	−0.004	0.97	0.09	0.47
Plasma IL-8	0.012	0.92	0.04	0.75

Analysis is performed using Spearman’s rank-correlation coefficient. Rho is the correlation coefficient; the absolute value can range from 0 to 1.

^*Tidal volume on enrollment adjusted for predicted body weight.

Higher mBAL total protein concentration was significantly associated with more ventilator-free days, fewer hospital days, and fewer ICU days. In unadjusted linear regression models, each 500 μg/ml increase in day 0 mBAL total protein was associated with an additional 0.8 VFDs (95% CI: 0.05–1.6, P = 0.038) (Table 3). Adjusting for age, serum protein, and vasopressor use did not attenuate this association, and it remained significant. A scatterplot of the raw data (Fig. 2) is consistent with the findings of the unadjusted linear model describing the relationship between increasing mBAL total protein concentration, and ventilator-free days. In unadjusted models, each 500 μg/ml increase in day 0 mBAL total protein was significantly associated with 0.6 fewer hospital days (P = 0.019) and 0.7 fewer ICU days (P = 0.018). Adjusting for age, serum protein, and vasopressor use, only slightly attenuated this association. The association between ventilator days among survivors and mBAL total protein was not significant in univariate or multivariate models. Higher mBAL IgM concentration was associated with more ventilator-free days and fewer ventilator days among survivors. In unadjusted linear regression models, each 50 ng/ml increase in mBAL IgM was associated with an additional 1.1 VFDs (95% CI: 0.2–2.1, P = 0.017). Among survivors, each 50 ng/ml increase in mBAL IgM was associated with fewer ventilator days, −1.0 (95% CI: 1.7 to −0.3, P = 0.008). Adjusting for age, serum protein and vasopressor use did not attenuate this association, and it remained significant. Although the point estimates describing the unadjusted and adjusted associations between mBAL IgM and hospital and ICU days among survivors were all negative, similar to the direction of the associations described above for mBAL total protein, they were not statistically significant.

Table 3.

Linear regression models testing the association of scaled mBAL total protein and IgM concentrations and clinical outcomes

	mBAL Total Protein β Coefficient for Each 500 µg/ml (95% CI)	mBAL IgM β Coefficient for Each 50 ng/ml (95% CI)
Ventilator-free days unadjusted	0.8 (0.05–1.6)	1.1 (0.2–2.1)
Ventilator-free days adjusted	0.8 (0.05–1.5)	1.1 (0.2–2.0)
Ventilator days unadjusted	−0.3 (−1.3–0.7)	−1.0 (−1.7 to −0.3)
Ventilator days adjusted	−0.4 (−1.0–0.1)	−1.0 (−1.7 to −0.4)
Hospital days unadjusted	−0.6 (−1.2 to −0.1)	−0.7 (−1.3–0.4)
Hospital days adjusted	−0.7 (−1.3 to −0.2)	−0.5 (−1.2–0.3)
ICU days unadjusted	−0.7 (−1.3 to −0.1)	−0.7 (−1.6–0.1)
ICU days adjusted	−0.8 (−1.4 to −0.2)	−0.6 (−1.3 to −0.1)

Adjusted analyses control for age, serum protein concentration, and vasopressor use in the 24 h preceding randomization. All analyses except ventilator-free days excluded subjects who died. All clinical endpoints refer to 28 days of observation. Significant associations are indicated in bold. CI, cofidence interval.

Fig. 2.

Scatterplot of day 0 mBAL total protein concentration and ventilator-free days to 28 days with best-fit line and linear prediction line for subjects with at least one ventilator free day (VFD). The best-fit and linear prediction lines exclude subjects with zero VFDs and the high zero-count of this outcome was accommodated with appropriate model checking.

In unadjusted logistic regression models, the concentrations of mBAL total protein and IgM were not significantly associated with mortality at 28 days (P value = 0.46 and 0.27, respectively). The findings were similar in logistic regression models adjusting for age, APACHE III score, and vasopressor use.

DISCUSSION

In both adjusted and unadjusted analyses, higher mBAL total protein concentration at day 0 was associated with more VFDs in patients with ARDS enrolled in the ALTA trial. These associations were independent of age, serum protein, and vasopressor use in the 24 h before enrollment. These findings were unexpected. We hypothesized that the direction of association would be similar to animal studies and to translational studies performed before the era of lung-protective ventilation, as well as one modest sized study of mBAL total protein in patients ventilated with low tidal volumes (3). Earlier studies have shown that protein concentration in BAL fluid obtained at later time points correlates with more severe alveolar injury and worse clinical outcomes (3, 5, 8, 10, 17, 21, 24, 29, 33). Although we observed weak correlations between mini-BAL total protein concentration and oxygenation index and $\mathrm{P}{\mathrm{a}}_{\mathrm{O}}{}_{{}_{\mathrm{2}}}$/$\mathrm{F}{\mathrm{I}}_{{\mathrm{O}}_{\mathrm{2}}}$ approaching significance, there was no significant correlation between mini-BAL total protein concentration and driving pressure, plasma IL-6, or plasma IL-8, markers of severity of lung injury and predictors of poor clinical outcomes after ARDS (4, 4a, 34). Adjustment for vasopressor use, a marker for severity of illness, did not attenuate the association between mBAL total protein. Furthermore, adjustment for APACHE III score at study enrollment did not improve the multivariate models. Therefore, the unexpected direction of the associations we observed does not appear to be related to these baseline measurements of severity of illness, oxygenation impairment, or pulmonary dysfunction.

When interpreting our findings in the context of the earlier literature, it is important to consider the dynamic nature of the alveolar fluid composition. The protein concentration in alveolar fluid reflects not only the flooding during acute injury, but also the mechanisms of alveolar fluid clearance through active transport, and the slower clearance of proteins through paracellular routes (11, 12). Thus, higher mBAL protein concentrations may identify ARDS patients with intact alveolar epithelium and preserved function of active transport mechanisms for the resolution of alveolar edema, resulting in higher protein concentrations and faster resolution of lung injury (Fig. 3). Although experimental data suggest that increasing alveolar fluid clearance with β-agonists can accelerate the resolution of alveolar edema, we observed no difference in mBAL total protein and IgM concentrations between the treatment arms of the ALTA trial (23, 25). It is possible that the small sample size of this study led to a type II error. Alternatively, perhaps the effect of the β-agonists on alveolar edema clearance is dependent on the mode of delivery. Injured alveoli may not take up inhaled β-agonist. Nebulized delivery of albuterol may not be as efficacious as methods used in other studies, including intravenous infusion or direct lobar installation of β-agonists.

Fig. 3.

Alveolar fluid composition is dynamic, reflecting not only the flooding during acute injury, but also alveolar fluid and protein clearance. This schematic is provided to illustrate a plausible biological mechanism that may explain our results. Intact alveolar units with active transport of sodium clearing edema fluid and effectively decreasing the volume of solvent. Because protein clearance is a slower process dependent on paracellular routes, the edema fluid clearance leads to a more concentrated alveolar edema fluid measured through alveolar lavage.

In a previously published secondary analysis of a randomized controlled trial of 75 subjects testing APC for acute lung injury (ALI) defined by American-European Consensus Conference criteria, higher mBAL total protein was correlated with fewer ventilator-free days (3, 6). This correlation was weak (rho = −0.25), but statistically significant, and was the opposite direction of our findings. When interpreting our findings in the context of these previously reported findings, it is important to consider differences in the study design and data analysis plans. Subjects were enrolled in ALTA within 48 h of meeting ALI criteria, while subjects were enrolled in the APC study within 72 h of meeting ALI criteria. The APC study excluded patients with sepsis and APACHE II score ≥25, and the cohort had a lower mortality rate (11%) than the subjects in the ALTA substudy (19%). Compared with the ALTA substudy, in the APC study, there was a lower incidence of vasopressor use before enrollment (24% vs. 41%). The APC study excluded trauma patients, while in this ALTA substudy study, nine (11%) of the subjects had trauma as the primary or secondary etiology of lung injury. Beyond these differences in study populations, the previously reported results from the secondary analysis of the APC study reported only the simple correlation coefficient between mBAL total protein and VFDs, and adjusted analyses were not performed.

The association of IgM concentration in day 0 mBAL with VFDs was similar to the association observed between total protein concentration in day 0 mBAL and VFDs. Our findings suggest that the observed association between protein concentration in the mBAL and the resolution of injury is independent of the size of proteins accumulating in the alveolar compartment. These data do not support the hypothesis that the extravasation of larger proteins into the alveolar compartment indicates more severely injured alveoli that may take longer to resolve. Instead, we found that in this cohort of patients with ARDS, the concentration of a large protein, IgM, in alveolar fluid did not provide additional prognostic information beyond the associations observed for mBAL total protein concentration.

This study has several strengths. To our knowledge, this is one of only two studies of alveolar protein concentration and resolution of lung injury in patients with early ARDS receiving low tidal volume ventilation (3). Because the mBAL samples were obtained as part of the ARDS Network ALTA study, there is extensive clinical data available on the enrolled subjects. Studying biomarkers in mBAL samples is appealing because mBAL is a safe procedure that is readily available at many institutions and can be performed by respiratory therapists.

This study has some limitations. This is a modest-sized, retrospective cohort study, and we cannot rule out the possibility of a type II error leading to the conclusion that there is no association between mBAL IgM or total protein and mortality. Because mBAL was only carried out on a subset of the larger cohort, we cannot eliminate the possibility of selection bias driving the association between mBAL biomarkers and VFDs. Because of exclusion criteria for safety, this substudy included patients with less severe oxygenation impairments. Only three patients (4%) had severe ARDS, as defined by an enrollment $\mathrm{P}{\mathrm{a}}_{\mathrm{O}}{}_{{}_{\mathrm{2}}}$/$\mathrm{F}{\mathrm{I}}_{{\mathrm{O}}_{\mathrm{2}}}$ <100, and most patients (n = 40, 51%) had mild ARDS with an enrollment $\mathrm{P}{\mathrm{a}}_{\mathrm{O}}{}_{{}_{\mathrm{2}}}$/$\mathrm{F}{\mathrm{I}}_{{\mathrm{O}}_{\mathrm{2}}}$ of > 200. Furthermore, the mean APACHE III score in this substudy was lower than the intervention and control arms of the larger ALTA study population (86 ± 22 vs. 94 ± 29 and 92 ± 30, respectively). Therefore, our findings may not be generalizable to a population with more severe lung injury or more severe underlying illnesses. It is also worth noting that this study used mBAL, and both animal studies and all but one of the earlier clinical studies have analyzed edema fluid protein concentration or BAL. We cannot accurately account for the effects of dilution or return yield in our calculations. However, it should be noted that we would expect any sampling variability introduced by mBAL to bias our findings toward the null. There is also a possibility of confounding that we were unable to identify. Another potential limitation of this study stems from statistical issues related to linear regression models and the distribution of the outcome of interest, ventilator-free days. Our choice of linear regression models to analyze the association between biomarkers and ventilator-free days requires special attention to checking the assumptions of the linear model. Although this outcome has a high zero count and is not normally distributed, we performed extensive model checking, including the bootstrap approach to confidence intervals to try to mitigate model misspecification. Although we considered other modeling choices, we chose this statistical approach because it is consistent with current critical care literature and familiar to clinicians (1, 16, 20, 22, 26, 28, 30). Although other nonparametric models (27), such as a Cox proportional hazards model, could have been used for this analysis, we believe that the established precedent in the literature and adequate model checking in our methodological approach supports our choice of linear regression analysis to test the association between mBAL total protein and VFDs.

In summary, in this retrospective study of patients with ARDS, higher early mBAL protein concentrations were associated with faster resolution of lung injury, resulting in shorter duration of mechanical ventilation and shorter length of stay in the hospital. Results for analyses of mBAL concentration of IgM were similar to mBAL total protein concentration; thus, measuring a large protein that is not normally secreted in the alveolar compartment did not substantively change the direction or magnitude of association between the mBAL biomarker of interest and measures of resolution of lung injury or clinical outcomes. Higher mBAL protein concentration may identify ARDS patients with intact alveolar epithelium and preserved function of active transport mechanisms for the resolution of alveolar edema, resulting in higher protein concentrations and faster resolution of lung injury. Our results suggest that future studies using mBAL in ARDS patients could be designed to obtain an early and a repeat measurement of protein concentration from the distal airspaces, which may indicate intact alveolar fluid clearance that may predict faster resolution of lung injury and improved clinical outcomes. Additionally, in future studies, simultaneously measuring other clinical markers of improvement along with serial mBAL protein concentrations could be most informative.

GRANTS

Dr. Hendrickson was supported by National Heart, Lung, and Blood Institute (NHLBI) Training Grants T32HL-007185 and F32HL-124911, and NHLBI K23HL-133495. Dr. Matthay was supported in part by NHLBI R37HL-51856 and R01HL-51854. Dr. Calfee was supported in part by HL-133390, HL-131621, and HL-110969. Dr. Liu was supported in part by NHLBI R37HL-51856.

DISCLOSURES

No conflicts of interest, financial or otherwise are declared by the authors.

AUTHOR CONTRIBUTIONS

C.M.H. and J.A. performed experiments; C.M.H., H.Z., K.D.L., and C.S.C. analyzed data; C.M.H., H.Z., K.D.L., C.S.C., and M.A.M. interpreted results of experiments; C.M.H. prepared figures; C.M.H. drafted manuscript; C.M.H., J.A., K.D.L., C.S.C., and M.A.M. edited and revised manuscript; C.M.H., J.A., H.Z., K.D.L., C.S.C., and M.A.M. approved final version of manuscript. NHLBI ARDS Network was responsible for study design and distribution of biospecimens for experiments.