INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

A fundamental problem with any form of investigative pulmonary lavage is that the quantity of epithelial lining fluid (ELF) washed out of the lung in the return fluid is unknown and variable. This uncertainty significantly limits the interpretation of measured solute concentrations, and necessitates the use of a marker substance to indicate the degree to which ELF has been diluted in the return fluid (1). The requirement for a marker of dilution applies regardless of whether the fluid is collected by fiberoptic bronchoalveolar lavage (BAL), or by the various forms of small volume lung lavage (SVLL) commonly employed in the ventilated patient (see companion article on pp. 771-777).

In the analysis of SVLL fluid collected from ventilated infants with lung disease, a number of different endogenous dilutional markers have been used (2), but their relative validity has not been fully assessed, and several issues require clarification (16). For dilutional markers such as total protein (2, 3), albumin (4), sphingomyelin (SM) (3, 7, 8), and IgA secretory component (SC) (9), the inherent assumption that ELF concentration is constant in all study subjects has not been adequately scrutinized. Several investigators have noted higher concentrations of protein (3, 12, 13), albumin (9, 10, 17), and SM (3) in SVLL fluid from infants with lung disease than from control infants, suggesting that it may be incorrect to assume constant ELF concentration for these compounds. Watts and Bruce (17) showed in tracheal aspirate (TA) fluid that lavage concentration of albumin, but not SC, was higher in diseased than in normal lungs. They concluded that albumin concentration in ELF ([Alb]_ELF) was variable in the study groups, but that [SC]_ELF was constant, and thus suitable as an endogenous dilutional marker. This conclusion may well be correct, but it ignores the fact that the amount of ELF recovered during lung lavage is not uniform, and high [Alb]_LAVAGE in the lung disease group may in part be explained by increased ELF volume in the return fluid. Clearly, definitive analysis of the validity of dilutional markers where constant ELF concentration is assumed requires that there be an independent estimate of ELF recovery, so that [Marker]_ELF can be calculated and compared in the groups under study.

As with adult BAL, urea has also been used as an endogenous dilutional marker in SVLL fluid analysis in ventilated infants (12). Although there are concerns that diffusion of urea into the lavage fluid from the pulmonary interstitium may occur during the lavage (18), the rapidity with which SVLL is performed in ventilated infants means that significant overestimation of ELF recovery due to urea influx is much less likely than with adult BAL (21). The possibility remains, however, that the amount of urea influx during SVLL may vary depending on the state of the pulmonary epithelium, such that ELF recovery will be selectively overestimated in patients with lung disease. This phenomenon has been observed in adults with lung disease undergoing fiberoptic BAL (22); whether it occurs in infants during SVLL is unknown.

In view of the uncertainties regarding the use of dilutional markers in SVLL, we studied concentrations of putative dilutional markers in lung lavage fluid taken from a group of ventilated infants with a variety of lung diseases, and in nonventilated control infants with normal lungs. The aims of this investigation were: (1) to examine concentrations of protein, albumin, SM, and SC, both in lavage fluid and in ELF, noting any differences between diseased and normal lungs, or any change in concentration over time in the diseased lung; (2) to compare SC content in lavage fluid from the infected lung, the noninfected diseased lung, and the normal lung; and (3) to examine the relationship between lavage/plasma urea ratio and indices of lung injury and dysfunction.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Infants under a year of age with lung disease requiring mechanical ventilatory support were eligible for enrollment in this study. Elucidation of the cause of the respiratory failure was achieved primarily using clinical and radiographic data; where relevant, respiratory pathogens were identified by bacterial culture of blood and TA, and/or viral immunofluorescence of nasopharyngeal washings. Healthy infants with normal lungs undergoing anesthesia for surgery were enrolled as controls. Infants with a prior history of any respiratory illness, and those undergoing cardiothoracic surgery, were deemed unsuitable as controls. Informed parental consent was obtained in all cases, and the study protocol was approved by our institutional Ethics in Human Research Committee.

Lavage fluid samples were collected from the ventilated infants in place of an episode of routine endotracheal suction. Repeated samples, at least 1 d apart, were taken in many infants during the course of their disease. For each sampling episode, ventilator settings, and results of most recent arterial blood gas analysis were recorded, and alveolar-arterial oxygen difference (AaDO₂) and oxygenation index (OI) were calculated: OI = [(mean airway pressure × FI_O2)/Pa_O2] × 100. The plasma urea estimation closest to the time of lavage was noted. Samples were taken from control infants immediately after intubation under anesthesia, prior to commencing surgery.

Paired lavage samples using both TA and nonbronchoscopic BAL (NB-BAL) sampling methods were collected where possible. NB-BAL samples were not taken in infants with marked asymmetry of disease (e.g., diaphragmatic hernia), nor in infants whose cardiorespiratory status was unstable. Time constraints in the operating room precluded TA and NB-BAL samples being performed contemporaneously in the control infants.

All samples were collected by a single operator (P.A.D.), with every effort made to standardize the sampling procedure. The methods used are fully described in the companion article. In brief, TA samples were collected using four aliquots of 0.5 ml saline, with each aliquot instilled directly down the endotracheal tube, followed by tracheal suction using a standard suction catheter. The catheter was then washed with 1 ml saline. For collection of NB-BAL samples, a straight end-hole suction catheter (No. 565; Vygon, Ecouen, France) was advanced down the endotracheal tube via a suction bullet and swivel adaptor, and gently wedged in the airway. Three 1 ml/kg aliquots of warmed saline were then instilled via the catheter, with low-pressure suction after each aliquot. Positive pressure ventilation continued throughout the lavage, which was completed in 60 to 90 s in all cases.

The specimens obtained were collected in a mucus trap, centrifuged at 150 × g and 4° C for 10 min, and stored at 20° C for later analysis. Blood staining of the specimen was quantified using the same 0-3 scale as described in the accompanying article. Samples that were heavily contaminated by mucus, or showed significant macroscopic blood staining, were discarded.

The following ELF constituents were assayed in the lavage fluid: disaturated phosphatidylcholine (DSPC) (23); surfactant protein A (SP-A) (24); SM (25); and total protein (26). SC was assayed with an enzyme immunoassay as previously described (27), using two commercially available antibodies to SC. Purified secretory IgA (sIgA) was used as the standard, and thus sample concentrations of free and bound SC were expressed as µg/ml equivalent of sIgA standard. Urea concentration in lavage fluid was measured by urease assay, and albumin by immunoturbidimetry, both using an automated analyzer.

For the purposes of comparison of the ELF concentration of protein, albumin, SM, and SC, a direct estimate of ELF recovery was made using urea (18), according to the formula: ELF volume (per ml of return fluid) = ([urea]_LAVAGE/[urea]_PLASMA). The ELF concentration of all analytes was calculated (e.g., [Prot]_ELF), as were the ratios of DSPC, SP-A, and albumin to SC.

Data Analysis

Logarithmic transformation of all data sets pertaining to lavage fluid constituents was undertaken. Geometric means were compared by t test or one-way analysis of variance (ANOVA) as appropriate. Correlation of lavage/plasma urea ratio with markers of lung injury and dysfunction was quantified using Spearman's rank correlation. For both DSPC and SP-A, multiple linear regression of surfactant content, expressed as a concentration in ELF, and as a ratio to SC (i.e., two predictor variables) was performed to determine which of the dilutional markers (urea or SC) best standardized the concentration of the surfactant index to optimize prediction of OI (the outcome variable).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Demographic and Sampling Data

A total of 90 TA and 61 NB-BAL samples were taken in the ventilated infants with lung disease. The "initial" sample in each of the infants (taken at a time of significant lung disease) was used for the analysis of the validity of protein, albumin, SM, and SC as dilutional markers. Table 1 shows the demographic data at the time of the initial TA and NB-BAL sampling in the ventilated infants, and in the control infants, in whom TA or NB-BAL (but not both) was performed. For the initial TA samples, the plasma specimen for urea estimation was drawn at a mean (± SD) of 4.7 ± 3.8 h from the time of lavage in the ventilated infants, and 3.1 ± 2.6 h in the control infants; for NB-BAL samples the respective time differences were 5.1 ± 3.3 and 4.4 ± 3.3 h. Mean (± SD) OI and AaDO₂ at the time of initial TA sampling in the ventilated group were 9.7 ± 8.8 and 270 ± 170, respectively, corresponding values for the NB-BAL group were 10 ± 12 and 290 ± 170. 

From the total pool of TA and NB-BAL samples, a subgroup of ventilated infants was identified in whom the initial sample was taken at a time of significant lung dysfunction (defined as AaDO₂ > 200), and a "recovery" sample was taken once improvement in lung function had been noted (AaDO₂ < 200). This group allowed comparison of dilutional marker concentrations as lung disease abated, and was representative of the whole patient population in terms of demographics and disease profile. Mean OI and AaDO₂ in the TA recovery samples (n = 17) were 4.4 ± 1.6 and 140 ± 33; corresponding values in the NB-BAL group (n = 10) were 4.2 ± 2.9 and 140 ± 35. The initial and recovery samples were 72 and 71 h apart in the TA and NB-BAL groups, respectively.

In infants with normal lungs, the range of raw concentration of surfactant constituents in lavage fluid was as follows: TA samples: DSPC 6.5-fold range in values, SP-A 24-fold; NB-BAL samples: DSPC 21-fold range, SP-A 91-fold range.

Concentration of Dilutional Markers in Lavage Return Fluid

Raw values and calculated ELF concentration of all dilutional markers in the diseased, recovering, and normal lung are shown in Table 2. In both TA and NB-BAL samples, raw concentrations of protein, albumin, and SM were greater in the initial sample from the lung disease group compared with control groups. Additionally, in TA fluid, [Alb]_ELF and [SM]_ELF were higher in lung disease than in control infants; these differences were less prominent in NB-BAL fluid. Raw concentrations of protein, albumin, and SM remained high in the TA recovery samples, but for NB-BAL were significantly lower than the initial values. [Prot]_ELF and [Alb]_ELF also decreased in the recovery NB-BAL samples. With both modes of sampling, ELF volume was not different between the groups although a trend toward higher ELF recovery in the initial NB-BAL samples was noted in infants with lung disease compared with control infants (p = 0.068). Raw and ELF concentrations of SC were not different between the initial, recovery, and control samples with either sampling technique.

Validity of SC as a Dilutional Marker in Infectious Lung Diseases

Whereas the SC content appeared not to differ in the diseased and normal lung overall, Table 3 shows that there was significant variation in both raw and ELF concentration of SC in specific disease groups. For NB-BAL samples, highest values for SC and [SC]_ELF were noted in the acute viral bronchiolitis (AVB) and pneumonia subgroups; a similar trend was noted in the pneumonia group in TA samples. Figure 1 further examines SC content in infective lung disease, showing individual data points for [SC]_ELF in controls, AVB, pneumonia, and a pooled noninfective group. In NB-BAL but not TA samples, [SC]_ELF was higher in AVB than in noninfective lung disease or group B streptococcal (GBS) pneumonia, and also trended higher than in control infants (p = 0.084). Of note in infants with pneumonia is that, both for TA and NB-BAL, the highest [SC]_ELF values occurred in patients with pathogens other than group B streptococcus. Additionally, values of [SC]_ELF in the NB-BAL samples were lower in the pooled noninfective group than in control infants; a similar trend was noted in the TA samples which just failed to reach statistical significance.

View larger version (22K): [in this window] [in a new window]   Figure 1.   [SC]ELF in initial samples from control infants, infants with noninfective lung disease, AVB, and pneumonia due to GBS and other pathogens. Horizontal bar = geometric mean. Lower than AVB, p < 0.05, t test; p < 0.01. Control values higher than noninfective group: TA: 160 versus 85 µg/ml, p = 0.051; NB-BAL: 32 versus 11 µg/ml, p = 0.024. Cont = control; Non-inf. = noninfective lung disease (hyaline membrane disease [HMD], meconium aspiration syndrome [MAS] ± congenital diaphragmatic hernia [CDH]); GBS = group B streptococcal pneumonia; RSV = respiratory syncytial virus. Other abbreviations as in Table 1.

Relationship between Lavage/Plasma Urea Ratio and Indices of Lung Injury and Dysfunction

To investigate whether influx of urea into lavage fluid was greater in the diseased lung with damaged epithelium, the strength of linear association between lavage/plasma urea ratio and several markers of severity of lung injury and dysfunction was quantified (Figure 2). No clear relationship with degree of blood staining was evident, although a trend toward a higher urea ratio in more heavily blood-contaminated NB-BAL specimens was seen. A weak positive correlation was noted between urea ratio and albumin/SC ratio (alb/SC) in NB-BAL samples, whereas for TA a weak negative association was observed. With both modes of sampling, no significant relationship was noted between urea ratio and OI, indicating that the relative amount of urea in lavage return fluid appeared to be largely independent of oxygen and ventilation requirements. No correlation was noted between [urea]_LAVAGE/ [urea]_PLASMA in contemporaneous paired TA and NB-BAL samples (r = 0.066, p = 0.64), further suggesting that even in severe lung injury there was no consistent additional influx of urea related to disruption of the pulmonary epithelium.

View larger version (35K): [in this window] [in a new window]   Figure 2.   Association between lavage/plasma urea ratio and markers of lung injury, all samples. Panels A, C, E: TA samples. Panels B, D, F: NB-BAL samples. Panels A, B: Natural log values of urea ratio (×1,000) versus degree of blood staining. Horizontal bar = mean of ln values. TA: No difference between blood staining groups, p = 0.89 (ANOVA). NB-BAL: No significant difference between groups, p = 0.18. Panels C, D: Log of urea ratio versus log of alb/SC. TA: Spearman's r = 0.24, p = 0.027; NB-BAL: Spearman's r = 0.32, p = 0.013. Panels E, F: Log of urea ratio versus log of OI. TA: Spearman's r = 0.038, p = 0.72; NB-BAL: Spearman's r = 0.14, p = 0.28.

Comparison of Urea and SC as Dilutional Markers

The relative validity of urea and SC were compared by determining which of these markers best standardized the raw concentration of surfactant indices in lavage fluid so as to predict lung dysfunction as measured by OI. In TA samples, multiple linear regression analysis did not find either urea or SC to be clearly superior as a dilutional marker (data not shown). For NB-BAL analysis, however, estimation of ELF recovery from the lavage/plasma urea ratio did appear to best standardize surfactant concentration to predict OI. Both [DSPC]_ELF and [SP-A]_ELF were independent predictors of OI in the linear regression equation (p = 0.028 and 0.0056, respectively), whereas DSPC/SC and SP-A/SC were not. Addition of the raw concentrations of DSPC and SP-A did not enhance prediction of OI in any of the regression equations.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The Need for a Dilutional Marker in SVLL

The data from this study first of all reaffirm that a dilutional marker is indeed necessary in the analysis of SVLL fluid to take account of the significant variation in ELF recovery that occurs during a single episode of lavage. This was manifest in the wide spread of values for lavage concentration of ELF constituents such as DSPC and SP-A, in keeping with the observations of Ratjen and coworkers (28), who found a 480-fold range in SP-A concentration in the normal lung using 3 × 1 ml/kg fiberoptic BAL. The direct estimates of ELF recovery made in our study using lavage/plasma urea ratio also varied widely, with a trend in the NB-BAL samples toward higher recovery in the diseased lung as has been noted previously (14). We also found, as did Hallman and coworkers (12), that prediction of lung dysfunction by surfactant indices in lavage fluid was enhanced by standardization of the raw concentration using a dilutional marker. Our data thus appear to confirm that the variability of ELF recovery inherent in any method of SVLL mandates that a dilutional marker be used to allow meaningful interpretation of solute concentrations in the lavage return fluid.

Validity of Protein, Albumin, SM and SC as Dilutional Markers

In the present study, the estimated ELF concentration of a number of candidate dilutional markers was examined in SVLL samples taken from normal and diseased infant lungs. Protein, albumin, and SM were found not to be valid dilutional markers, as their ELF concentrations differed either between the normal and diseased lung, or between the initial and recovery sample in the diseased lung, or both. These differences were distributed relatively evenly between TA and NB-BAL samples (Table 2). In contrast, ELF concentrations of SC did not differ in SVLL fluid from the normal and diseased lung, nor did they vary significantly between the initial and recovery samples in the lung disease group. As has been suggested previously (10, 17), SC certainly appears to be a more convincing candidate as a marker of dilution in SVLL fluid than protein, albumin, or SM.

For the NB-BAL samples, the lavage fluid concentrations of protein, albumin, and SM in our study agree well with those of previous investigations that used similar lavage volume. In lavage fluid collected from the normal lung by 3 × 1 ml/kg fiberoptic BAL, Ratjen and coworkers (28) and Riedler and coworkers (29) found protein concentrations of 100 and 130 µg/ ml, respectively, compared with our value of 130 µg/ml. Albumin values were also similar in the samples of Riedler and coworkers (29) and our own (49.8 µg/ml versus 35 µg/ml). By contrast, in TA fluid from control infants, LeVine and coworkers (3) found a higher protein and SM concentrations than in our study (protein 5,100 versus 540 µg/ml, SM 8.1 versus 1.9 µg/ml), but the infants they enrolled as controls were ventilated and postoperative, and may have had some exudative alveolar edema.

In our control infants, the estimated ELF concentrations of protein, albumin, and SM also closely correspond to previous estimates using similar lavage techniques. The value for [Prot]_ELF in TA fluid was very similar to that of Hallman and coworkers (12) and of Contreras and coworkers (13) (4.6, 6.4, and 5 mg/ml, respectively). Additionally, the TA and NB-BAL values for [Alb]_ELF in control infants (0.68 and 0.89 mg/ ml) are remarkably similar to those obtained by direct sampling of airway surface liquid from the normal lung (0.48 and 0.73 mg/ml in the trachea and distal airway, respectively) (30). Our value for [SM]_ELF in NB-BAL fluid (26 µg/ml) is somewhat lower than that obtained by extrapolation of the data of Ijsselstijn and coworkers (8) (63 µg/ml), where lavage fluid was collected by 2 × 1 ml/kg NB-BAL in 3- to 5-d-old ventilated control infants.

Several investigators have noted, as we did, differences in raw concentration of total protein (3), albumin (9, 10, 17), and SM (3) in SVLL fluid from the normal and diseased infant lung. Our study data suggest, however, that these findings are partly explained by increased ELF recovery, in that the marked differences in lavage fluid concentration of protein, albumin, and SM in the diseased, recovering, and normal lung were in many cases diminished when their respective ELF concentrations were compared. This finding supports our view that realistic conclusions about validity of dilutional markers in lavage fluid cannot be drawn unless an estimate of their concentration in ELF has been made.

Clearly, in the comparison of ELF concentration of protein, albumin, SM, and SC in this study, the assumption has been made that the estimate of ELF recovery using lavage/ plasma urea ratio is accurate, with negligible influx of urea into lavage fluid during the procedure. During fiberoptic BAL in adults, significant urea influx does occur (18), but it has been suggested that if lavage is conducted rapidly, with a dwell time of 60 s or less, estimates of ELF recovery using urea will be sufficiently accurate to justify use of the method (21). The data of Grigg and coworkers (31), who found no increase in lavage/plasma urea ratio in sequential analysis of return fluid from a 2 × 1 ml/kg NB-BAL, support the view that urea influx can be minimized during SVLL. Similarly, in TA fluid from tracheotomized fetal rabbits, Hallman and coworkers were unable to demonstrate any overestimation of ELF recovery related to consistent urea influx (12), again suggesting that urea can be a reliable dilutional marker in SVLL fluid.

The comparison of [Marker]_ELF performed in this study also assumes that, if there is an influx of urea leading to overestimation of ELF volume, the error is equivalent in all samples. This may not be the case; it has been suggested that urea influx will be greater in the damaged epithelium of the diseased lung (22), although Ward and coworkers (20) were not able to reproduce these findings. In the comparisons we made, in each case where a difference in [Marker]_ELF was noted, the more severely diseased lungs had the higher value. If greater urea influx was selectively occurring in the diseased state, this would only have the effect of narrowing the difference observed in [Marker]_ELF, so that the real differences between ELF concentration of protein, albumin, and SM may in fact be higher than is indicated in Table 2 and are very unlikely to be lower.

Validity of SC as a Dilutional Marker in the Presence of Lung Infection and Lung Injury

In the comparison between the diseased, recovering, and normal lung, raw and ELF concentrations of SC did not differ overall, confirming its potential as a valid dilutional marker. Differences were noted, however, in [SC]_ELF in different disease groups, with the highest values in NB-BAL fluid from infants with AVB and non-GBS pneumonia. These data suggest that SC should not be used as a dilutional marker where the lung disease is infective in nature. This conclusion is at odds with that of Watts and Bruce (17), who found no change in [SC]_LAVAGE in TA samples from a cohort of 10 preterm infants with systemic sepsis and respiratory deterioration. It is not clear from their data whether these infants had confirmed bacterial pneumonia, and augmented epithelial production of SC in response to pathogenic organisms is unlikely to occur in the absence of localized infection. Additionally, as has been discussed, without an estimate of ELF recovery it is not possible to know whether [SC]_ELF varied between their groups, even if [SC]_LAVAGE did not.

Our data also revealed that [SC]_ELF was lower in a pooled noninfective lung disease group than in control infants, and raises the possibility that nonspecific injury to the tracheobronchial tree leads to a decrease in secretion of SC into the airway. This mechanism has been previously suggested to explain the lower concentration of SC in BAL fluid noted in adult smokers (32), and in asthmatics (33). It is possible, therefore, that using SC as a dilutional marker may somewhat underestimate ELF recovery in the ventilated infant with noninfective lung disease, particularly in NB-BAL samples.

Validity of the Lavage/Plasma Urea Ratio in Estimation of ELF Recovery in SVLL

In assessing the validity of urea as a dilutional marker in this study, our primary aim was to determine whether changes in respiratory epithelial permeability in the diseased lung were associated with an increase in urea influx into the lavage fluid, thereby selectively overestimating ELF volume. In the TA samples, no clear relationship was found between markers of lung injury and [urea]_LAVAGE/[urea]_PLASMA, other than a weak negative association with alb/SC. For NB-BAL, a weak positive correlation was noted between urea ratio and alb/SC, but not degree of blood staining or OI. We used alb/SC as a simple index of the integrity of the respiratory epithelium (33), in that it relates the concentration of a protein exclusively derived from plasma to that of a lung-specific protein with a very low plasma concentration (17). The association between urea ratio and alb/SC in NB-BAL samples suggests that in the presence of increased protein exudation, either the recovery of ELF is greater, or it is overestimated because of increased urea influx during the lavage. We, like others (22), have no firm data that allow us to distinguish between these two possibilities. However, the absence of a relationship between urea ratio and blood staining or OI, the negative correlation with alb/SC observed in TA samples, and the lack of correlation of urea ratio values in paired TA and BAL samples, weigh against there being a major additional influx of urea related to heightened epithelial permeability in the diseased lung.

Comparative Validity of Urea and SC as Dilutional Markers

The multiple linear regression analysis used in this study assumes that even in our heterogeneous study population, alveolar surfactant status will significantly affect pulmonary function, and thus that a linear association between surfactant indices and OI should exist, as it does in infants with hyaline membrane disease (HMD) (12). Our data suggest that for NB-BAL samples, urea may be a better dilutional marker to standardize solute concentrations than SC, whereas in TA samples no clear differentiation can be made. This finding is not surprising given the virtual absence of SC secretion in the distal airspaces (34), as evidenced by the marked difference in SC concentration found in contemporaneous paired TA and NB-BAL samples (see accompanying article).

In conclusion, we have shown by analysis of SVLL fluid that (1) protein, albumin, and SM are not appropriate dilutional markers as their ELF concentration varies significantly in the diseased, recovering, and normal lung; (2) SC may be a valid dilutional marker, but should not be used in the context of lung infection; and (3) lavage/plasma urea ratio appears to provide a meaningful estimate of ELF recovery, with no clear evidence for selective overestimation of ELF volume in the diseased lung related to greater urea influx into the lavage fluid. We would propose that for TA samples, SC or urea are satisfactory dilutional markers, whereas for NB-BAL fluid, urea is a more appropriate choice.