Published ahead of print on September 27, 2007, doi:10.1164/rccm.200705-655OC
© 2008 American Thoracic Society doi: 10.1164/rccm.200705-655OC
Bronchoalveolar Lavage and Response to Cyclophosphamide in Scleroderma Interstitial Lung Disease1 Division of Pulmonary and Critical Care Medicine and 2 Division of Rheumatology and Immunology, Department of Medicine, Medical University of South Carolina, Charleston, South Carolina; 3 Division of Pulmonary and Critical Care Medicine, Department of Medicine, David Geffen School of Medicine at UCLA, Los Angeles, California; 4 Division of Pulmonary, Allergy, and Critical Care Medicine, Department of Medicine, Boston University School of Medicine, Boston, Massachusetts; and 5 Department of Radiology, 6 Division of Rheumatology and Arthritis, Department of Medicine, and 7 Department of Biostatistics and Biomathematics, David Geffen School of Medicine at UCLA, Los Angeles, California Correspondence and requests for reprints should be addressed to Charlie Strange, M.D., Division of Pulmonary and Critical Care Medicine, MUSC, 96 Jonathan Lucas Street, 812 CSB, Charleston, SC 29425. E-mail: strangec{at}musc.edu
Rationale: The presence of inflammatory cells on bronchoalveolar lavage is often used to predict disease activity and the need for therapy in systemic sclerosis–associated interstitial lung disease. Objectives: To evaluate whether lavage cellularity identifies distinct subsets of disease and/or predicts cyclophosphamide responsiveness. Methods: Patients underwent baseline lavage and/or high-resolution computed tomography as part of a randomized placebo-controlled trial of cyclophosphamide versus placebo (Scleroderma Lung Study) to determine the effect of therapy on forced vital capacity. Patients with 3% or greater polymorphonuclear and/or 2% or greater eosinophilic leukocytes on lavage and/or ground-glass opacification on computed tomography were eligible for enrollment. Measurements and Main Results: Lavage was performed in 201 individuals, including 141 of the 158 randomized patients. Abnormal cellularity was present in 101 of these cases (71.6%) and defined a population with a higher percentage of men (P = 0.04), more severe lung function, including a worse forced vital capacity (P = 0.003), worse total lung capacity (P = 0.005) and diffusing capacity of the lung for carbon monoxide (P = 0.004), more extensive ground-glass opacity (P = 0.005), and more extensive fibrosis in the right middle lobe (P = 0.005). Despite these relationships, the presence or absence of an abnormal cell differential was not an independent predictor of disease progression or response to cyclophosphamide at 1 year (P = not significant). Conclusions: The presence of an abnormal lavage in the Scleroderma Lung Study defined patients with more advanced interstitial lung disease but added no additional value to physiologic and computed tomography findings as a predictor of progression or treatment response. Clinical trial registered with www.clinicaltrials.gov (NCT 000004563).
Key Words: systemic sclerosis • Scleroderma Lung Study
Interstitial lung disease (ILD) in systemic sclerosis (SSc) is common and remains the leading cause of death (1–3). Considerable heterogeneity exists in the severity and progression of lung disease. However, a subset of patients with SSc ILD follows a course of progressive decline (4–6). In these patients, inflammatory changes in the lung occur early in the course (7–9) and progress over time (10, 11). Although lung parenchymal abnormalities may be absent on chest computed tomography (CT) at the earliest stage of ILD, most patients develop ground-glass opacities (GGO) and parenchymal lung distortion. Open lung biopsy in these individuals usually demonstrates either cellular or fibrotic nonspecific interstitial pneumonitis (NSIP) (12, 13). These observations have suggested that immunosuppressive therapy for active inflammation with cyclophosphamide (CYC) may limit the development of lung fibrosis (6, 14–22). The randomized, placebo-controlled Scleroderma Lung Study (SLS), which demonstrated a beneficial effect of oral CYC in slowing progression of restrictive lung disease and improving dyspnea and health-related quality-of-life measures, confirmed these observations (23). Bronchoalveolar lavage (BAL) has been used in SSc to define a population of patients whose lung function will likely decline without therapy (4, 6, 10, 14). The best correlates of subsequent decline in forced vital capacity (FVC) have included the percentage of polymorphonuclear leukocytes (PMNs) and eosinophils on a standardized BAL from either the right middle lobe (RML) or lingula (4, 14, 24). However, these cellular components of BAL have also been seen in other ILDs, such as idiopathic NSIP, idiopathic pulmonary fibrosis (25), and cryptogenic organizing pneumonia (26). In these diseases, BAL cell counts and cellular differentials have been variably associated with disease severity (25, 27), but in most studies these do not predict subsequent disease progression (28, 29). High-resolution CT of the chest (chest HRCT) is able to detect and characterize abnormalities associated with scleroderma lung disease (30–34). In the SLS, the baseline chest HRCT fibrosis score was the strongest correlate of disease progression in those assigned to the placebo arm and of the difference between 1-year FVC in the CYC versus placebo groups (23). Fibrosis of the lung produces airway distortion and traction bronchiectasis that is often associated with abnormalities of mucus clearance and excess BAL neutrophils (35). In this context, prior associations between BAL cellularity and outcome may have represented an epiphenomenon that would have been better characterized by the extent of fibrosis on chest CT, particularly since chest HRCT is universally available and less invasive than BAL. This article reports the BAL characteristics in the SLS cohort and explores the utility of BAL with or without chest HRCT to define a CYC-responsive SSc population. Some of the results of these studies have been previously reported in the form of abstracts (36–38).
Patient Selection The SLS is a multicenter, randomized, double-blind, parallel comparison of CYC versus placebo in dyspneic SSc patients with ILD that enrolled patients between September 2000 and January 2004. All subjects provided written, informed consent. Eligible subjects had onset of their first non-Raynaud's manifestation within 7 years and an FVC between 45 and 85% predicted. All subjects had been nonsmokers for at least 6 months. Complete inclusion and exclusion criteria appear in Table E1 of the online supplement. Patients meeting initial screening criteria were invited to undergo HRCT and BAL, and those with GGO on HRCT and/or a positive BAL (  3.0% neutrophils and/or 2.0% eosinophils) were randomized at a 1:1 ratio to receive CYC ( 2 mg/kg/d) or matching placebo in a double-blind fashion for 1 year. Pulmonary function was assessed every 3 months during the first year to determine the primary endpoint of %predicted FVC at 12 months, adjusted for baseline. Full methods have been published previously (23).
BAL
Pulmonary Function Testing
Chest HRCT
Statistical Analysis
Identification of Two BAL Cohorts A total of 267 subjects were screened for eligibility at 13 clinical centers throughout the United States. Of these, 201 underwent BAL without a serious adverse event (Figure 1). A total of 158 study-eligible subjects were randomized, of whom 141 had undergone both an evaluable BAL and chest HRCT. These 141 individuals make up the population examined in this article.
A positive BAL was observed in 101 (71.6%) of these subjects and an abnormal chest HRCT demonstrating any form of GGO was present in 125 subjects (88.7%). The distribution of cellular findings on BAL is shown in Figure E1. Two cohorts, 101 subjects with a positive BAL and 40 subjects with a negative BAL (no increase in BAL neutrophils or eosinophils), were defined for further analysis (Table 1). The screening HRCT had no GGO or fibrosis in 15.8% of individuals. Twelve subjects also had a BAL lymphocytosis of greater than 10%, but there was no overall difference in the distribution of lymphocytes between the BAL-positive (BAL+) and BAL-negative (BAL–) groups and no specific correlation of lymphocytosis with HRCT findings or CYC responsiveness.
Comparison of Cell Differentials between Local and Core Readers In general, cell differential counts performed at the clinical centers correlated well with those performed at the core laboratory with respect to both the percentage of neutrophils (  coefficient, 0.72; 95% confidence interval [CI], 0.60–0.84) and eosinophils ( coefficient, 0.60; CI, 0.46–0.74) and with the participant's assignment as either BAL+ or BAL– according to the entry criteria ( coefficient, 0.75; CI, 0.63–0.88). In the absence of the core reading, none of the 101 BAL+ subjects would have been ruled ineligible by these criteria and 5 of the 40 BAL– subjects would have been identified as belonging to the BAL+ cohort. However, with the combined use of BAL and/or HRCT as entry criteria, all 141 of the subjects still would have been enrolled into the trial regardless of whether the BAL differential was performed at the clinical center or the BAL core.
Baseline Characteristics of the BAL+ and BAL– Cohorts
Univariate correlates of % PMNs and % eosinophils similar to the global BAL+ correlates are presented in the online supplement. A cough producing sputum was correlated with % PMNs but not % eosinophils. The volume of BAL recovered from the RML averaged 115.4 ± 41.3 ml (48% of instilled volume) for all subjects, but was significantly greater in the BAL– subset than the BAL+ subset (131.6 ± 35.6 vs. 109.1 ± 41.8 ml, P = 0.003) (Table 1). The nature of this difference, and the cause for the wide intersubject variability in BAL return (range, 26–229 ml) was investigated by univariate correlations to a variety of patient variables. Significant relationships identified by this approach were then examined by a multiple regression technique (Table 4). Prior cigarette smoking, TLC %predicted, FEV1/FVC, and the presence of 3% or more PMNs were identified as independent correlates of BAL volume.
Relationship between BAL Findings and SLS Outcomes BAL cellular differentials were examined to identify relationships between the presence and extent of abnormality and the change in baseline-adjusted FVC over the first 12 months of treatment. Only 126 of the 158 randomized subjects had both baseline BAL results and met the primary analysis criteria of completing at least 3 months of therapy and having follow-up FVC measurements available at 6, 9, and/or 12 months. Within this defined cohort, assignment to the placebo arm was associated with a significant decline in %predicted FVC over time (Figure 2B; P = 0.0001, linear mixed model). When examined with respect to the findings on BAL, there was not a significant difference in the course of subjects who were BAL+ versus those who were BAL– (P = 0.73, linear mixed model). Even when corrected for both the baseline FVC and the severity of fibrosis (worst fibrosis score) on HRCT (23), which were worse in those with a positive BAL, there were no detectable differences in the change over time of the %predicted FVC (P = 0.12) between the BAL+ and BAL– groups. Consistent with this result, a multiple regression analysis failed to identify any specific BAL finding (including total cellularity, total lavage volume returned, or the percentage of PMNs, lymphocytes, or eosinophils) as an independent predictor of the primary outcome.
Although those assigned to CYC had a significantly better baseline adjusted %predicted FVC at 12 months than those on placebo in the primary analysis for the SLS (23), this difference between treatment groups no longer reached a statistically significant level when the analysis was limited to the 126 patients with evaluable BAL cell differentials (P = 0.075). We therefore examined the impact of BAL-defined cohorts on the proportion of subjects who had either a stable or improving %predicted FVC at 12 months versus the proportion of subjects with a worsening %predicted FVC (Figure 3). A significantly higher proportion of those assigned to receive CYC had a stable or improving FVC compared with placebo (P = 0.015, Fisher's exact test). When divided into the BAL+ and BAL– cohorts, a significant treatment effect for CYC was only observed in those who were in the BAL+ (P = 0.034) and not in the BAL– cohort (P = 0.42).
A small but unique subpopulation included 14 individuals who enrolled with a positive BAL but normal HRCT for whom 1-year FVC data were available (see Figure E2). There was no difference in the percentage of individuals who improved with CYC (3 of 6; 50%) compared with the individuals who received placebo (4 of 8; 50%), although the very small size of this subpopulation could lead to misleading results. However, there also was no difference compared with the larger population with a positive BAL and any GGO on HRCT who were treated with CYC (21 of 40; 52.5%) (P = not significant).
The poor prognosis associated with SSc ILD, its variable response to therapy, and the toxicities associated with treatment emphasize the importance of identifying patients who are more likely to have progressive disease and/or to respond to treatment. Historically, progressive shortness of breath, declining lung function tests, and/or a changing radiograph have triggered clinicians to initiate therapy. In 1993, Silver and colleagues reported that FVC stabilized or improved when a group of 14 dyspneic patients with SSc and with a declining FVC and alveolitis on their BAL were treated with oral CYC (16). In a retrospective case study, White and coworkers also reported a strong correlation between inflammatory changes on BAL ( 3% neutrophils and/or 2% eosinophils) and the course of FVC and DLCO over time depending on whether or not the patients were treated with CYC (14). Patients with SSc and ILD and a negative BAL had, on average, stable values for FVC and DLCO over the mean observation period of 16 months. In contrast, those with a positive BAL who were not treated experienced a significant decline in lung function, whereas those with a positive BAL who were treated with CYC experienced an improvement in their FVC and DLCO over time. With the assumption that inflammation causes progressive fibrosis and that CYC mediates its effects by reducing lung inflammatory cell numbers (41), the measurement of inflammatory cells on BAL has become a common practice for identifying patients with SSc and ILD who warrant immunosuppression. However, controversy has remained about the reproducibility of BAL cell differentials in general and the utility of a single RML lavage as a measure of lung inflammatory cells (43), and the relationship between BAL cellularity and disease activity (11). With respect to this issue, the SLS provided a unique opportunity to prospectively study the relationship between BAL cell differentials and the progression of SSc ILD in a randomized controlled trial in which patients were treated with either oral CYC or placebo and carefully monitored over the course of 1 year. There were several features of the SLS that were designed to reduce bias related to the interpretation of the BAL cell differentials. First, a combination of chest HRCT and BAL entry criteria was used in a redundant manner to define patients with active ILD. As such, the study population contained a mixture of BAL+ and BAL– patients, with the BAL– cohort representing 28% of those enrolled. In addition, the cell differentials were read independently by two out of three experienced readers and the results presented as the average interpretation, thereby reducing individual selection bias. All discordant interpretations, which would have resulted in a different eligibility assignment, were resolved by consensus. Finally, BAL cell differentials were read at both the clinical center and at the core to evaluate the feasibility and reproducibility of BAL readings in the clinical setting. Our results demonstrate a high degree of correlation between reading sites and that, by using both HRCT and BAL as entry criteria, there was little chance for differences between readers to have influenced the outcome. As such, the SLS demonstrated that a multicenter study using BAL can be performed with minimal variance between centers, if appropriate efforts to standardize the procedure are used. The first goal of our analysis was to determine whether the presence of a positive BAL identified a unique cohort of patients with SSc and ILD with respect to their baseline features or the rate of progression of their disease. Indeed, the BAL+ cohort had significantly more severe lung disease by pulmonary function criteria, more pure GGO on HRCT, and a trend toward more extensive fibrosis. In fact, when findings on the BAL were correlated to a regional analysis of the RML on HRCT, the most significant difference between the BAL+ and BAL– groups was the severity of fibrosis, which was worse in those with a positive BAL. A similar relationship between fibrosis and a positive BAL was recently reported in a smaller study of SSc ILD by Clements and colleagues (43). For some years, idiopathic pulmonary fibrosis was believed to be an inflammatory disease on the basis of BAL studies, until it was later recognized that usual interstitial pneumonia on open lung biopsy does not have as many inflammatory cells as were seen on BAL. One explanation is that lung fibrosis causes traction of small airways that leads to mucus retention and cellularity in the small airways. Whether NSIP associated with SSc shares this quality remains unknown. The results of this study would suggest that BAL cellularity may have little to do with pathogenesis and is simply an epiphenomenon. Therefore, the "alveolitis" that has historically been characterized as the lesion of SSc ILD may not be of alveolar origin at all. Alternatively, GGO likely represents a mixture of microscopic fibrosis, interstitial cellularity, and excesses of alveolar cells that characterize NSIP. The difference in baseline lung disease and fibrosis between the BAL+ and BAL– subsets also raises concern about the uncontrolled comparison of these two groups in prior studies—which assumed that they were equal in all other aspects. A multivariate analysis from the main SLS identified two independent factors that were related to the severity of lung function at the conclusion of the study: baseline FVC and baseline fibrosis as measured by HRCT (23). There was a significant and negative correlation between the baseline fibrosis score on HRCT and the 12-month value for %predicted FVC, suggesting that the presence of fibrosis is a predictor of progressive disease. As such, prior studies using BAL findings as a predictor of disease activity may have inadvertently selected for more extensive disease at baseline. In our current analysis, we evaluated the changes in %predicted FVC over time using both the measured values and after adjusting outcomes for the observed differences in baseline FVC and fibrosis, neither of which identified a difference in the course of untreated disease between those subjects who were BAL+ and those who were BAL–. In addition, past cigarette smoking that could impact BAL recovery or cellularity was not found to impact changes in %predicted FVC in either the CYC or the placebo group. The second goal of this analysis was to determine whether the presence of a positive BAL predicts a subset of patients more likely to respond to CYC treatment. The only baseline variables correlated with a differential treatment response as measured by the difference in baseline-adjusted %predicted FVC at 12 months were the baseline FVC and the baseline HRCT fibrosis score. Neither BAL cellularity nor identification as BAL+ or BAL– by entry criteria predicted outcome when subjected to a multivariate analysis. Unfortunately, the SLS was not sufficiently powered to allow subgroup stratification with respect to its primary analysis comparing the CYC and placebo arms for any variables. Therefore, the analysis we performed in this study was an effort to define the degree of colinearity with baseline FVC and the baseline HRCT fibrosis score and determine if any statistical measure of BAL cellularity could be correlated with FVC change. Any correlation found would necessarily be less robust than the baseline FVC and the baseline HRCT fibrosis score for clinical care. We examined the impact of stratification by BAL subgroup on the proportion of subjects in whom the %predicted FVC was either stable or improving versus deteriorating. A significant treatment effect from CYC was observed in the BAL+ group, but not in the BAL– group. However, given the considerably smaller sample size of the BAL– group, it remains difficult to conclude that there is a true association between BAL cellularity on BAL and response to CYC. A potential limitation in assessing the utility of BAL in this study was that lavage was performed exclusively in the RML, regardless of the site of disease activity by chest HRCT. Some studies have suggested that interlobar variation of BAL cellularity is small in the majority of individuals (42); however, higher percentages of eosinophils and PMNs are found when the BAL is performed in an area of chest HRCT abnormality in idiopathic pulmonary fibrosis (35) and in SSc (43). As such, our approach may have underestimated the number of BAL+ subjects and reduced our ability to correlate BAL findings with outcomes. Future studies could consider multilobar BAL as an alternative sampling strategy, similar to the targeting of open lung biopsies from multiple areas of variably involved lung in ILD. The clinical importance of the current analysis is significant. If chest HRCT and pulmonary function testing are better predictors of a beneficial CYC response than BAL, then BAL should rarely be performed for clinical care in SSc unless infection is suspected (43, 44). If fibrosis is the important feature that drives ILD progression, a strong argument can be advanced that therapy for SSc ILD should be instituted when the chest HRCT shows fibrosis, regardless of BAL findings. Because fibrosis will thereafter be present on all subsequent CT scans, further clinical trials will be needed to determine optimal duration of treatment. In summary, despite the correlation of a positive BAL to measures of lung fibrosis and the severity of baseline lung function abnormality, BAL was not helpful in predicting either the course of disease in the placebo group or the response to CYC as measured by FVC %predicted. When subjected to the rigorous analysis of a randomized controlled trial, the clinical utility of BAL cellularity as a routine test to assess the activity of SSc ILD appears limited in the setting in which pulmonary function and HRCT can more readily be obtained. The utility of BAL for other biomarkers of disease activity or clinical response to therapy is not precluded by this study. Instead, BAL may remain a useful tool to exclude lung infections when clinically indicated, and to obtain cells and epithelial lining fluid for studies of SSc ILD pathogenesis.
The authors thank Michael Bonner, Clementine Singleton, and Katie Caldwell for their assistance with laboratory specimens.
Supported by Public Health Service grants from the National Heart, Lung, and Blood Institute, National Institute of Arthritis and Musculoskeletal and Skin Diseases, and National Center for Research Resources of the National Institutes of Health. Cyclophosphamide (Cytoxan) was supplied by Bristol-Myers Squibb.
* A complete list of members may be found at the end of this article. This article has an online supplement, which is accessible from this issue's table of contents at www.atsjournals.org Originally Published in Press as DOI: 10.1164/rccm.200705-655OC on September 27, 2007 Conflict of Interest Statement: None of the authors has a financial relationship with a commercial entity that has an interest in the subject of this manuscript. The Scleroderma Lung Study Research Group includes the following contributing members: University of California, Los Angeles, Los Angeles, CA (UO1 HL 60587 and UO1 HL 60606, for the Data Coordinating Center.): Philip J. Clements, M.D., M.P.H.; Donald P. Tashkin, M.D.; Robert Elashoff, Ph.D.; Jonathan Goldin, M.D., Ph.D.; Michael Roth, M.D.; Daniel Furst, M.D.; Ken Bulpitt, M.D.; Dinesh Khanna, M.D.; Wen-Ling Joanie Chung, M.P.H.; Sherrie Viasco, R.N.; Mildred Sterz, R.N., M.P.H.; Lovlette Woolcock; Xiaohong Yan, M.S.; Judy Ho; Sarinnapha Vasunilashorn; Irene da Costa. University of Medicine and Dentistry of New Jersey, New Brunswick, NJ (UO1 HL 60550): James R. Seibold, M.D.; David J. Riley, M.D.; Judith K. Amorosa, M.D.; Vivien M. Hsu, M.D.; Deborah A. McCloskey, B.S.N.; Julianne E. Wilson, R.N. University of Illinois, Chicago, Chicago, IL (UO1 HL 60895): John Varga, M.D.; Dean Schraufnagel, M.D.; Andrew Wilbur, M.D.; David Lapota, M.D.; Shiva Arami, M.D.; Patricia Cole-Saffold, M.S. Boston University, Boston, MA (UO1 HL 60682): Robert Simms, M.D.; Arthur Theodore, M.D.; Peter Clarke, M.D.; Joseph Korn, M.D.; Kimberley Tobin; Melynn Nuite, B.S.N. Medical University of South Carolina, Charleston, SC (UO1 HL 60750): Richard Silver, M.D.; Marcie Bolster, M.D.; Charlie Strange, M.D.; Steve Schabel, M.D.; Edwin Smith, M.D.; June Arnold; Katie Caldwell; Michael Bonner. Johns Hopkins School of Medicine, Baltimore, MD (UO1 HL 60597): Robert Wise, M.D.; Fred Wigley, M.D.; Barbara White, M.D.; Laura Hummers, M.D.; Mark Bohlman, M.D.; Albert Polito, M.D.; Gwen Leatherman, M.S.N.; Edrick Forbes, R.N.; Marie Daniel. Georgetown University, Washington, DC (UO1 HL 60794): Virginia Steen, M.D.; Charles Read, M.D.; Cirrelda Cooper, M.D.; Sean Wheaton, M.D.; Anise Carey; Adriana Ortiz. University of Texas, Houston, Houston, TX (UO1 HL 60839): Maureen Mayes, M.D., M.P.H.; Ed Parsley, D.O.; Sandra Oldham, M.D.; Tan Filemon, M.D.; Samantha Jordan, R.N.; Marilyn Perry. University of California, San Francisco, San Francisco, CA (UO1 HL 60587): Kari Connolly, M.D.; Jeffrey Golden, M.D.; Paul Wolters, M.D.; Richard Webb, M.D.; John Davis, M.D.; Christine Antolos; Carla Maynetto. University of Connecticut Health Center, Farmington, CT (UO1 HL 60587): Naomi Rothfield, M.D.; Mark Metersky, M.D.; Richard Cobb, M.D.; Macha Aberles, M.D.; Fran Ingenito, R.N.; Elena Breen. Wayne State University, Detroit, MI (UO1 HL 60839): Maureen Mayes, M.D.; Kamal Mubarak, M.D.; Jose L. Granda, M.D.; Joseph Silva, M.D.; Zora Injic, R.N., M.S.; Ronika Alexander, R.N. Virginia Mason Research Center, Seattle, WA (UO1 HL 60823): Daniel Furst, M.D.; Steven Springmeyer, M.D.; Steven Kirkland, M.D.; Jerry Molitor, M.D.; Richard Hinke, M.D.; Amanda Mondt, R.N. University of Alabama, Birmingham, AL (UO1 HL 60748): Mitchell Olman, M.D.; Barri Fessler, M.D.; Colleen Sanders, M.D.; Louis Heck, M.D.; Tina Parkhill. All centers listed above received support by a grant from the NHLBI. Data Safety and Monitoring Board members are as follows: Taylor Thompson, M.D. (Harvard Medical School, Boston, MA); Sharon Rounds, M.D. (VA Medical Center, Brown University, Providence, RI); Michael Weisman, M.D. (Cedars Sinai/UCLA, Los Angeles, CA); Bruce Thompson, Ph.D. (Clinical Trials Surveys, Baltimore, MD). Mortality and Morbidity Review Committee members are as follows: Harold Paulus, M.D. (UCLA); Steven Levy, M.D. (UCLA); Donald Martin, M.D. (Johns Hopkins University). Received in original form May 2, 2007; accepted in final form September 27, 2007
Related articles in AJRCCM:
This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
AT A GLANCE COMMENTARY
. Comparisons in BAL interpretation were assessed using the 
