INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The prognosis of lone cryptogenic fibrosing alveolitis (CFA; synonymous with idiopathic pulmonary fibrosis) remains poor; most patients do not respond to therapy, and only 50% survive more than 5 yr after the onset of symptoms (1). Fibrosing alveolitis is seen in the great majority of patients with systemic sclerosis (4, 5) and it is histologically identical to lone CFA (6); the CT features of the two diseases are also very similar (7). However, despite their histologic and radiologic similarities, CFA and FASSc have different courses; survival in FASSc is substantially better than in lone CFA, after controlling for age, sex, smoking history, treatment, histologic confirmation of the diagnosis, and disease severity (evaluated using physiologic and morphologic variables (2). Differences in markers of disease activity between these two diseases may explain differences in outcome and thus have the potential to give insights into the pathogenesis of fibrosing alveolitis. Bronchoalveolar lavage (BAL) cell counts have been widely used as markers of inflammation in the lower respiratory tract in fibrosing alveolitis. In CFA, a BAL neutrophilia and/or eosinophilia is associated with a worse outcome with treatment, whereas a BAL lymphocytosis identifies an increased likelihood of response to therapy (8).

The aims of this study were twofold. Firstly, BAL counts were compared between CFA and FASSc in untreated ex-smoking or nonsmoking patients, after adjustment for lobar disease severity on CT, in order to gain insights into possible pathogenetic differences between the two diseases. Secondly, BAL cell counts were examined against lobar and global disease severity using multivariate techniques in order to determine whether lavage findings were primarily related to the global morphologic and functional severity of disease or to regional (lobar) abnormalities.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients

Consecutive patients with CFA or FASSc undergoing BAL at the Royal Brompton National Heart and Lung Hospital between January 1986 and December 1990 were studied; exclusion criteria consisted of (1) current smokers within 6 mo of BAL; (2) patients receiving corticosteroid therapy, cyclophosphamide, or azathioprine.

Fifteen patients met clinical criteria for a diagnosis of CFA, consisting of (1) persistent bilateral crackles on auscultation; (2) a restrictive ventilatory defect or isolated depression of gas transfer on pulmonary function testing; (3) CT appearances compatible with fibrosing alveolitis; (4) the absence of environmental exposure to any known fibrogenic agent. The diagnosis of fibrosing alveolitis was confirmed by open lung biopsy in 28 of 47 patients (FASSc, n = 20; CFA, n = 8). Thirty-two patients met preliminary criteria of the American Rheumatism Association for a diagnosis of systemic sclerosis (11) and had appearances typical of fibrosing alveolitis on CT; BAL findings in 27 of these patients have been reported previously (12). In three patients, FVC and carbon-monoxide diffusing capacity (DL_CO) levels were normal, and three patients did not have persistent basal crackles (including one patient with normal FVC and DL_CO). Eight of 32 patients were receiving penecillamine (125 to 500 mg daily). Clinical details, extent of disease on CT, and functional data are shown in Table 1.

Pulmonary function tests, expressed as percentages of predicted values based on age, sex, height, and ethnic origin (13), were completed within 1 mo of CT scanning and BAL. FVC and DL_CO were analyzed. Spirometric measurements were made using a rolling-seal spirometer (Spiroflow; PK Morgan Ltd, Rainham, Kent, UK). DL_CO was measured by the single-breath technique, using transfer factor equipment (Model B; PK Morgan).

Bronchoscopy and Bronchoalveolar Lavage

Bronchoscopy with BAL was performed with the informed consent of the patient as part of the routine clinical management of CFA and FASSc. Patients with FEV₁ less than 1 L and PO₂ less than 70 mm Hg were excluded from BAL.

Premedication was administered intramuscularly (atropine, 0.6 mg; papaveretum, 5 to 10 mg) 30 min before BAL. Supplemental oxygen was administered during and immediately after the procedure. After the administration of topical 1% xylocaine, the bronchoscope was inserted transnasally and, after inspection of segmental orifices, was wedged according to a standard protocol in a subsegment of the right lower lobe, but using the middle lobe in some instances, depending upon technical factors. In three patients with more extensive disease in the left lung, as judged by chest radiography, the left lower lobe was lavaged; selection of the site of BAL was not based on CT appearances. Aliquots of 60 ml of sterile normal saline were instilled through the bronchoscope, and the fluid was retrieved by mechanical suction. The standard introduction volume was 240 ml (n = 35) or 300 ml (n = 2); in 10 of 47 patients, the introduction volume was limited to 180 ml because of patient intolerance. The mean (SD) fluid volume introduced was 230 (29) ml. The mean (SD) volume of fluid retrieved was 76 (24) ml, a percentage yield of 33.6% (11.5%).

Cells were harvested from BAL fluid by low speed centrifugation at 300 × g for 5 min at 4^o C and were washed three times with cold minimal essential medium (MEM) containing 25 mM Hepes buffer. BAL cell counts were completed as previously reported (14). Total cell counts were made, using an improved Neubauer counting chamber, and expressed as the total number of cells per milliliter of aspirated fluid. Slide preparations for differential percentage counting of cells were made in a Shandon cytocentrifuge (Cytospin II; Shandon Ltd, Runcorn, Cheshire, UK) using 100-µl aliquots of the lavage cell suspensions, adjusted to 1.25 × 10⁶ cells/ml in MEM. After fixation in methanol, the preparations were stained with May-Grunwald Giemsa stain. Differential counts were made from a minimum count of 300 cells.

Thin-section Computed Tomography

Lung CT was performed as part of a routine clinical protocol; all patients underwent CT less than 4 wk before BAL. Scans were performed on an Elscint 2002 scanner (Elscint, Haifa, Israel) with a scanning time of 5.5 s. Interspaced 3-mm sections were acquired at 10-mm intervals from the lung apices to the bases, with the patient supine; sections were reconstructed with a high resolution "bone" algorithm (level, 500 to 700; width, 600 to 1,000 Hounsefield units). In cases showing increased opacification in the posterobasal segments, a limited number of sections were also acquired through the lower zones of the lung with the patient prone, to ensure that opacification was not due to gravitational effects.

Extent of abnormal lung was scored globally and in the lavaged lobe by two independent observers. Each observer made a visual assessment of the percentage of abnormal lung (to within 5%) at the five levels (the origin of the great vessels, the midarch of the aorta, the main carina, the pulmonary venous confluence, and 1 cm above the right dome of the diaphragm). An estimate of overall lung involvement was made by summing the scores at each level. The mean estimate of the two observers was used in analysis. The semiquantitative method of quantifying extent of disease on CT has a low interobserver variability (15) and has been applied to studies of fibrosing alveolitis (2, 16) and other diffuse interstitial lung diseases (19, 20). Sections taken through the lavaged lobe were assessed independently by two observers without knowledge of BAL findings. The percentage of abnormal lung within the lavaged lobe was estimated to the nearest 5%. The mean of the two observations was used in analysis.

Global CT patterns, evaluated independently by both observers, were graded using a system based upon the observations of Muller and colleagues (21, 22) and validated against histologic findings at open lung biopsy in patients with FASSc (23). The relative extent of a reticular pattern, denoting fibrosis (21, 23, 24), and a ground-glass pattern, representing an increased likelihood of inflammation (22, 23), were scored as follows: Grade 1, ground-glass attenuation more extensive than reticular abnormalities; Grade 2, ground-glass attenuation and reticular abnormalities equally extensive; Grade 3, reticular abnormalities more extensive than ground-glass attenuation. For divergent observations, the two observers reached consensus agreement after a further review. The extent of reticular abnormalities and ground-glass attenuation within lavaged lobes was scored by consensus agreement; the relative proportion of abnormal lung attributable to reticular abnormalities and to ground-glass attenuation was quantified to the nearest 5%. From this figure, the overall extent of reticular abnormalities and, separately, of ground-glass attenuation were calculated.

Data Analysis

Data analysis was carried out using STATA software. Group comparisons were made using Student's t test or, when appropriate, chi-squared statistics. BAL cellularity was not normally distributed; the Mann-Whitney U test was used for unadjusted comparisons of BAL cell counts between CFA and FASSc. Spearman's rank correlation was used to evaluate univariate relationships between lavage cellularity and measures of disease severity. Interobserver variation for CT observations was quantified as single determination standard deviations (for extent of disease on CT) and as the weighted kappa coefficient of agreement (K_w) for global CT grade (25).

Prior to multivariate analysis, abnormally distributed variables (all cell counts, the extent of ground-glass attenuation within lavaged lobes) were subjected to zero-skewness logarithmic transformation. Multivariate linear regression models (26) were constructed to identify independent determinants of BAL cellularity. The validity of the assumptions of linear regression was confirmed using residual-versus-predictor plots, tests for omitted variables (none identified in any analysis, p > 0.20 for all regression equations) and tests for heteroscedasticity (p > 0.20 for all regression equations). To ensure that significant relationships could not be ascribed to regression outliers, equations were re-examined after the exclusion of the three observations generating the greatest equation leverage (identified using leverage-versus-residual plots).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Observer Variation for CT Observations

The single determination standard deviation was 5.7% for lobar extent of disease and 5.1% for global disease extent. Interobserver variation for the global grading of the predominant CT pattern (K_w, 0.59) was indicative of good but not excellent interobserver agreement.

Unadjusted Comparisons between CFA and FASSc

Patient details at the date of BAL are summarized in Table 1. Patients with CFA were older, more often male, and had more severe disease within the lavaged lobe and globally (as judged by extent of disease on CT and depression of DL_CO). Unadjusted differences in BAL cell counts are shown in Figures 1-3. Percentage neutrophil counts (p < 0.03, Figure 1), total neutrophil counts/ml (p < 0.02, not shown), percentage eosinophil counts (p < 0.02, Figure 2), and total eosinophil counts per ml (p < 0.002, not shown) were higher in CFA. Percentage lymphocyte counts (p < 0.02, Figure 3) were higher in FASSc; total lymphocyte counts/ml did not differ between the two diseases (p = 0.75).

View larger version (10K): [in this window] [in a new window]   Figure 1.   Neutrophil differential percentages in CFA and FASSc (with median values shown). Neutrophil percentages were higher in CFA, p < 0.03. The upper limit of normal (for healthy nonsmokers) is indicated by the dashed line.

View larger version (10K): [in this window] [in a new window]   Figure 2.   Eosinophil differential percentages in CFA and FASSc (with median values shown). Eosinophil percentages were higher in CFA, p < 0.02. The upper limit of normal (for healthy nonsmokers) is indicated by the dashed line.

View larger version (11K): [in this window] [in a new window]   Figure 3.   Lymphocyte differential percentages in CFA and FASSc (with median values shown). Lymphocyte percentages were higher in FASSc, p < 0.02. The upper limit of normal (for healthy nonsmokers) is indicated by the dashed line.

Unadjusted Relationships between BAL Cellularity and Measures of Disease Extent

As shown in Table 2, neutrophil percentages and absolute counts were more strongly positively related to measures of the extent of disease on CT (Figures 4-6) than to reduction in DL_CO. There were weak positive relationships between total cell counts and most measures of disease extent. Eosinophil absolute counts (but not percentage differentials) were weakly positively related to selected measures of disease extent. There were no significant relationships between BAL cellularity and FVC levels.

View larger version (14K): [in this window] [in a new window]   Figure 4.   Neutrophil differential percentages in relation to the global extent of disease. A positive relationship could be ascribed to the presence of more extensive disease within lavaged lobes; on multivariate analysis, neutrophil levels reflected lobar disease extent, with no independent relationship to the global extent of disease on CT.

View larger version (15K): [in this window] [in a new window]   Figure 5.   Neutrophil differential percentages in relation to the extent of disease on CT within the lavaged lobe.

View larger version (13K): [in this window] [in a new window]   Figure 6.   Neutrophil differential percentages in relation to the extent of a reticular pattern on CT within the lavaged lobe.

BAL Cellularity in Relation to the Type of Fibrosing Alveolitis and the Severity of Disease

Independent relationships were evaluated between BAL cellularity and (1) the type of fibrosing alveolitis (CFA versus FASSc), and (2) measures of disease severity (the extent of disease on CT within the lavaged lobe, the global extent of disease on CT, DL_CO, and FVC). The determinants of total cell counts, neutrophil levels, and eosinophil levels are shown in Tables 345.

1. Total cell counts/ml did not differ between CFA and FASSc in any analysis, but they increased with increasingly extensive disease on global CT (p = 0.04) and with greater impairment of DL_CO (p = 0.03) (Table 3).
2. Neutrophil levels did not differ between CFA and FASSc, except when FVC was used to control for disease severity, and they were largely unrelated to functional measures (FVC and DL_CO); by contrast, neutrophil percentages and absolute counts were strongly linked to the lobar extent of disease on CT (Table 4). Equations in which disease severity was controlled using the lobar or global extent of disease on CT had higher explanatory powers (R² = 0.23 to 0.36) than equations utilizing functional measures (R² = 0.07 to 0.18). In order to determine whether regional and global morphologic extent of disease were both independent determinants of BAL neutrophil levels, analyses were repeated with the inclusion of the type of fibrosing alveolitis, extent of disease on CT within the lavaged lobe, and the global CT extent of disease in the same equation; absolute and percentage neutrophil counts remained positively related to lobar disease extent (p < 0.005 and p = 0.01, respectively), but they were not independently related to global disease extent.
3. Eosinophil levels (numbers per milliliters and percentage counts) were significantly increased in CFA in all analyses, but they were not independently related to measures of lobar or global disease severity (Table 5).
4. Lymphocyte levels did not differ between CFA and FASSc, and they were unrelated to measures of disease severity (data not shown).

In order to ensure that relationships were robust when identical diagnostic criteria were used for CFA and FASSc, analyses of neutrophil and eosinophil levels were repeated with the exclusion of the five FASSc patients with normal lung function (FVC and DL_CO) levels and/or absence of bilateral crackles. Findings were unaltered for eosinophil levels (higher in CFA for absolute and percentage eosinophil counts, p < 0.05, irrespective of the method used to control for disease extent). Relationships involving neutrophil levels were also preserved, with the exception of the higher percentage neutrophil counts seen in CFA when FVC was used to control for disease severity (p = 0.09, nonsignificant). No new statistically significant relationships were disclosed.

BAL Cellularity in Relation to Lobar and Global CT Patterns

The total cell count was unrelated to the extent of individual lobar CT patterns, but it increased with increasingly reticular global CT appearances (p = 0.05) (Table 6). Neutrophil levels were strongly positively linked to the extent of a reticular pattern in the lavaged lobe (p < 0.0005) and increased with increasingly reticular global CT appearances (p < 0.01). There was a weak relationship between eosinophil levels and the extent of a ground-glass pattern in the lavaged lobe (eosinophil %, p = 0.05; eosinophil/ml, p = 0.08) (Table 7). Lymphocyte levels were unrelated to lobar or global CT appearances (data not shown).

The Effect of Controlling for Demographic and BAL Technical Factors

In order to evaluate the robustness of independent relationships involving neutrophils and eosinophils, further linear regression models were constructed, incorporating as covariates the type of fibrosing alveolitis (CFA versus FASSc), lobar extent of disease, lobar site (middle lobe versus lobe), volume of fluid instilled at BAL, fluid percentage yield, age, sex, and smoking history (ex-smokers versus lifelong nonsmokers). Total eosinophil counts per milliliter and eosinophil percentage counts were independently increased in CFA (p < 0.005, equation R² = 0.28 and p = 0.01, equation R² = 0.23, respectively), but they were not independently related to other covariates. Total neutrophil counts per milliliter and neutrophil percentage counts increased with increasingly extensive lobar disease (p < 0.001, equation R² = 0.32 and p < 0.0005, equation R² = 0.31, respectively), but they were not independently related to other covariates.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The aim of this study was to evaluate relationships between BAL cellularity and the extent and type of fibrosing alveolitis. BAL neutrophil levels were more strongly associated with extensive fibrotic disease within the lavaged lobe than with the global extent of disease, evaluated functionally and morphologically (using CT), and they differed little between CFA and FASSc after adjustment for the extent of disease. By contrast, BAL eosinophil levels were consistently higher in CFA than in FASSc, but they were not independently related to disease extent.

The Importance of Controlling for Disease Extent in BAL Studies

In fibrosing alveolitis, the poor outcome associated with a BAL neutrophilia or eosinophilia (8) does not necessarily indicate that these cells are central to pathogenesis. It has been proposed that the neutrophil might play a key role in fibrosing alveolitis in the early generation of lung injury through oxidant and proteolytic enzyme release (27, 28). However, others have argued that the influx of neutrophils that might occur late in the course of disease, when disease is already advanced and fibrotic (29), does not reflect the underlying alveolitis (30) and may have little part to play in early pathogenesis. If this question is to be resolved, underlying disease extent must be taken into account when BAL findings are examined in relation to survival: it can be argued that BAL components predictive of survival, independent of disease extent, are more likely to be related to pathogenetic mechanisms. However, before controlling for the amount of disease in future prognostic analyses, it is necessary to determine which functional or morphologic measures should be used and which components of BAL cellularity are linked to the underlying extent of disease.

The present study has demonstrated that neutrophil levels are linked to the morphologic extent of disease in fibrosing alveolitis. This finding suggests that adjustment for disease extent is warranted in future prognostic studies of BAL neutrophil levels, but it may be less essential in studies of BAL eosinophil and lymphocyte levels. The neutrophil was the only cell to be strongly independently related to the extent of disease; weaker univariate relationships between eosinophils and the extent of disease were ascribable, on multivariate analysis, to more extensive disease in patients with CFA. Our findings indicate that the method of controlling for disease severity is likely to be important. Neutrophil levels were more strongly linked to the extent of disease within the lavaged lobe than to global disease extent on CT, and they were poorly related to functional indices; equations utilizing CT measures of disease extent had much higher explanatory powers (R² values) for neutrophil levels than did equations containing FVC or DL_CO levels. Furthermore, differences in neutrophil levels between CFA and FASSc were entirely accounted for by the fact that patients with CFA had more extensive fibrosing alveolitis and greater depression in DL_CO. By contrast, the use of FVC to control for disease extent did not entirely correct differences in neutrophil levels between CFA and FASSc, illustrating the importance of selection of the method to control for disease severity. Thus, in future studies of the relationship between BAL cellularity and survival, adjustment for the extent of disease on CT appears to be warranted; in previous BAL prognostic analyses, disease severity has not been controlled for (31), or functional indices have been used for this purpose (3).

The precision with which CT can be used to control for disease extent is dependent upon technical and methodologic factors. The use of 3-mm CT collimation in the present study, rather than the optimal section width of 2 mm, might have weakened relationships between disease extent and BAL cellularity to a minor degree. However, these limitations applied equally to patients with CFA or FASSc and should not have influenced comparisons between the two diseases; moreover, there is little difference between 1-mm and 3-mm collimation in the depiction of fine parenchymal abnormalities (32). Similarly, the use of a subjective semiquantitative means of estimating the extent of disease on CT can be debated, but the use of the objective measurement of disease extent in fibrosing alveolitis, using density masking, has not been validated, and the use of subjective scoring was modeled on that used in earlier clinical studies (2, 16).

Pathogenetic Implications

Comparisons between CFA and less progressive variants of fibrosing alveolitis (such as FASSc) have the potential to provide indirect evidence of possible pathogenetic mechanisms. In the present study, the higher BAL eosinophil levels in CFA, independent of disease extent, suggests that an eosinophilic influx may be linked to the inexorable progression of disease seen in CFA. Previously, eosinophil numbers and eosinophilic cationic protein concentrations in BAL fluid have correlated with reduced diffusing capacity in CFA (33). Poor responsiveness to therapy in patients with CFA with an initial BAL eosinophilia is well recognized (8); Turner-Warwick and colleagues (34) found that resolution of BAL eosinophilia at follow-up lavage was confined to the minority of patients responding to prednisolone or combination prednisolone/cyclophosphamide therapy. Schwartz and colleagues (35) reported that marked decline in DL_CO in CFA was associated with higher concentrations of eosinophils, but not neutrophils, in BAL fluid. Fujimoto and colleagues (36) categorized the clinical course of patients with fibrosing alveolitis into chronic stable, progressive, and acute progressive subgroups; patients with rapid clinical and physiologic decline were characterized by strikingly higher levels of BAL eosinophils and eosinophilic cationic protein. Recently, Boomars and colleagues (31) reported a correlation between high eosinophil levels and mortality, in the only study to examine survival in CFA from the date of BAL; neutrophil levels did not predict survival. Concentrations of BAL eosinophil degradation products have been found to correlate with functional impairment in FASSc (37, 38), and extracellular accumulation of major basic protein has been demonstrated in diseased lung tissue (38); increases in BAL eosinophil levels have been observed in established FASSc (39, 40). Taken together, these reports and the present study lend indirect support for further appraisal of the role of the eosinophil in the pathogenesis of fibrosing alveolitis. In particular, it is not clear whether the eosinophil plays an important pathogenetic role, or whether a BAL eosinophilia is merely a by-product of a central pathogenetic mechanism and does not contribute directly to disease progression. This question lies beyond the scope of the present study and can only be addressed further by tissue-based evaluation of inflammatory cells and mediators.

By contrast, the identical neutrophil levels in CFA and FASSc, after adjustment for the extent of disease on CT, make it difficult to argue that the excess mortality in CFA is ascribable to the toxic effects of neutrophil degradation products. The striking relationship between neutrophil levels and the extent of a reticular pattern (denoting fibrosis) within the lavaged lobe mirrors earlier findings (in a subgroup of the patients with FASSc reported in the present study) (12) and suggests that the major influx of neutrophils may occur when disease is already extensive and fibrotic.

The Problem of Ensuring Disease Comparability

One difficulty in comparing two variants of fibrosing alveolitis, CFA and FASSc, is to ensure that differences are disease-specific and do not result from disparities in disease severity at presentation. In systemic sclerosis, abnormalities evident on routine clinical evaluation and heightened physician awareness of the possibility of lung involvement lower the diagnostic threshold for fibrosing alveolitis, with the result that many patients with FASSc in the present study had limited disease. The higher BAL eosinophil levels seen in CFA persisted when inclusion criteria were standardized. However, the exclusion of five patients with FASSc without auscultatory crackles and/ or functional impairment (who did not meet diagnostic criteria for CFA) did not, in itself, correct functional and morphologic differences between the two diseases. The extent of disease on CT was chosen as the cardinal measure to control for disease severity because of the close similarity in CT appearances between CFA and FASSc. Importantly, in a recent comparative study of 52 patients with FASSc and 55 patients with CFA, there was little difference between the two diseases in the distribution of disease on CT or the coarseness of a reticular pattern, once the extent of disease on CT had been taken into account (7). Thus, the use of the extent of disease on CT to control for morphologic severity effectively adjusts also for other CT morphologic differences between CFA and FASSc. However, as no single functional or morphologic measure perfectly captures disease extent, functional indices were used as alternative measures of disease severity in separate models. Multivariate analysis involves the examination of multiple relationships and sometimes generates statistically significant findings through chance alone, which are lost when further analyses are performed using alternative covariates. However, the higher BAL eosinophil levels in CFA in the present study were robust in all analyses, irrespective of the means used to quantify disease extent.

It remains possible that tertiary populations provide a particular subset of patients with CFA with progressive disease who come to medical attention by virtue of progressive morbidity, whereas patients with FASSc include the full spectrum of patients with progressive and nonprogressive disease. Even after rigorous adjustment for disease severity, BAL differences between CFA and FASSc at tertiary institutions might not be disease-specific but might instead represent selection biasses caused by the differing presentations of the two disorders. This problem might be overcome by matching also for the rate of disease progression in the two patient groups, but in that case, BAL differences could not be evaluated as a source of pathogenetic hypotheses. In the present study, it can be argued that differences identified on BAL for a given disease extent are associated with differences between rapidly progressive and slowly progressive fibrosing alveolitis, whether because of selection by disease or because of selection by mode of presentation.

In conclusion, BAL eosinophil levels were higher in CFA than in FASSc, a less progressive disease, and this lends further support to the accumulated evidence that the eosinophil may be linked to disease progression in fibrosing alveolitis, either directly or as a marker of a central pathogenetic process. Neutrophil levels were identical in the two diseases after matching for disease extent, suggesting that the neutrophil has an equivalent pathogenetic role in both disorders and is not a determinant of the markedly worse outcome in CFA. BAL neutrophil levels were related more strongly to the extent of disease on CT, especially the extent of fibrosis within the lavaged lobe, than to functional measures, indicating the potential importance of CT as a measure of disease extent in future BAL studies.