INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The acute respiratory distress syndrome (ARDS) is the clinical manifestation of bilateral and severe diffuse alveolar damage (DAD) after an acute precipitating insult (1). Although the pathogenesis of DAD and its repair and progression through a fibroproliferative phase remain poorly understood, loss of integrity of the delicate alveolocapillary barrier is an early characteristic feature. Loss of this integrity results in pulmonary edema in the absence of an elevated pulmonary capillary pressure, and a marked increase in the protein concentration of the epithelial lining fluid (ELF). This is a dynamic process, and resolution of alveolocapillary permeability is critical to both the recovery of lung function and survival (2).

The alveolocapillary barrier is extraordinarily thin, ~ 0.1 to 0.2 µm, with a surface area of 50 to 100 m². Because the endothelial pore size is 6.5 to 7.5 nm, and the epithelial pore size is almost one tenth that at 0.5 to 0.9 nm, the epithelium is the major barrier to protein flux (3). Consequently, in the healthy lung a gradation in protein concentration is formed from the plasma to the interstitium to the ELF. The ELF probably has a protein concentration of ~ 20 g/L, 60 to 80% of which are low M_r proteins in equilibrium with plasma. In ARDS, the edema fluid:plasma protein ratio is typically greater than 0.8 (4). Locally secreted proteins, particularly the surfactant-associated proteins (SP), account for the remaining proteins in the ELF. Because protein flux across the alveolocapillary barrier is bidirectional, it is not surprising that we (5, 6) and others (7, 8) have reported elevated plasma levels of SP in ARDS.

Biologic markers have attracted a lot of attention in both acute lung injury (ALI) and ARDS (9) since they may allow (1) insights into the pathogenesis and pathophysiology of lung injury, (2) prediction of ARDS in at-risk patients, and (3) prediction of outcome. Because DAD must precede the development of ARDS, and because a loss of integrity of the alveolocapillary membrane is an early manifestation of DAD, we hypothesized that elevated SP in plasma would predict which patients would develop ARDS.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This study was approved by the Flinders Medical Centre Committee for Clinical Investigation (Permit No.26/93), and informed consent was obtained from the subjects or their closest relatives.

Patient Selection

Fifty-four consecutive, consenting patients admitted to the critical care unit and receiving respiratory support with either noninvasive ventilation or intubated ventilation (lung injury score [LIS] < 2.5) (10) for reasons other than left ventricular failure were recruited to the study. A risk factor for the development of ARDS must have been identified and plasma and matching physiologic data for the calculation of the LIS collected within 8 h. Risk factors were categorized as defined by Gregory and coworkers (11); however, three patients with pre-eclampsia, lung contusion, and pulmonary embolism were included under "other risk factor." A direct lung insult was categorized when the insult was pneumonia, aspiration of gastric contents, lung contusion, and pulmonary embolism, and the remaining risk factors were categorized as indirect lung insults. The clinical criteria for ARDS were bilateral diffuse lung infiltrates on chest radiograph, and a LIS  2.5 (10), not caused by left ventricular failure. After recruitment, plasma was collected for a further 3 d (Days 1 to 3), the LIS was collected daily until free from ventilatory support, and each patient was followed to hospital discharge. Ventilator-free days were defined as the number of days (up to Day 28) when the patient was free from mechanical ventilation (extubated or breathing without continuous positive airway pressure or pressure support) for more than 24 h.

Blood was drawn from an indwelling arterial catheter, immediately centrifuged in lithium heparin tubes at 5,000 rpm for 5 min at room temperature (Megafuge; Heraeus-Christ, Osterode, Germany), and the plasma was stored at 20° C for batch analysis. Samples were assayed in duplicate at four serial dilutions in a blind, randomized manner for SP-A and SP-B, using our previously described ELISA method (5, 6). The treating clinician was not aware of these data.

Measurement of SP-A and SP-B

Both SP-A and SP-B undergo extensive post-translational modification. Whereas the primary translation product of SP-A has a M_r ~ 28 kD, SP-A more correctly refers to a family of related thiol- and non-thiol-dependent glycoproteins, which, at least in humans, are derived from at least two gene products. In the air spaces SP-A is highly, and variably, glycosylated and associates as an oligomer comprising 18 peptides, M_r ~ 650 kD. In the case of SP-B, the protein is first synthesized as a precursor, which, after signal peptide cleavage, probably in the endoplasmic reticulum after translocation, and glycosylation, yields a hydrophilic proprotein (M_r ~ 42 kD). It is generally accepted that processing of the proprotein involves at least two distinct proteolytic events. The first, cleavage of 200 amino acids at the N-terminus to afford a ~ 25 kD intermediate, and the second, cleavage of 102 amino acids at the C-terminus to yield the mature protein. Although lamellar bodies, the major secretory source of surfactant lipids, are associated with mature SP-B but little, if any, SP-A or SP-B proprotein or ~ 25 kD intermediate, numerous quandaries remain regarding the secretion and processing of both SP-A and SP-B.

We have validated our assays using other independently sourced monoclonal (5) and polyclonal (6) antibodies. In the bloodstream we (5) and others (12) have shown that SP-A complexes with IgG and IgM. On the other hand, the immunochemical staining pattern for immunoreactive SP-B in plasma is similar to that in whole lung lavage and corresponds to reactivities with nominal M_r ~ 42 to 45 kD and ~ 24 to ~ 26 kD. Our blotting analysis fails to detect mature SP-B in plasma. However, this may only reflect the difficulties associated in separating, blotting, and detecting the hydrophobic protein against the sea of other more hydrophilic plasma constituents.

Although the exact nature of the SP-B antigens in blood remains to be clarified, we have recently analyzed plasma from an infant independently confirmed by two laboratories to have an autosomal recessive frame shift mutation at 121 base pairs in the SP-B gene, which resulted in a premature stop codon such that no SP-B mRNA or protein was produced. This definitive control has confirmed that our assay does indeed measure immunoreactive SP-B (Unpublished data).

Statistics

Data were analysed using SPSS for Windows Release 9.01 and AccuRoc 1.2 (13) for nonparametric receiver-operating characteristic (ROC) analysis. Because the SP-A and SP-B data were not normally distributed (Kolmogorov-Smirnov test with Lilliefors significance correction, p = 0.004 and p < 0.001), data are presented as medians with 25 to 75% quartiles. The Mann-Whitney U test and chi-square analysis was used to analyze differences between groups, and nonparametric ROC analyses were used to determine the optimal threshold values by maximizing the product of sensitivity and specificity. Spearman's nonparametric correlations were used to examine relationships between plasma SP-A or SP-B and ventilator-free days, hospital mortality, and plasma creatinine.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Outcome

Twenty of the 54 patients (41%) developed ARDS, 17 of them by Day 1 (Table 1). No patient developed ARDS from Day 4 to the end of the study. Age, sex, and diagnosis were evenly distributed between those who developed ARDS and those who did not. When the diagnoses were subdivided into direct and indirect lung insults, more patients with direct lung injury developed ARDS than did those with an indirect lung insult. Consequently, patients with direct and indirect lung insults were analyzed separately (Table 2). There was no difference in the time required to develop ARDS (1 d [1-1] versus 1 d [1-1], p = 0.94), or the time to the maximum LIS (1 d [1-2] versus 2 d [1-3], p = 0.31), between the direct and indirect lung insult groups, respectively.

At study entry the LIS was not different between those who subsequently developed ARDS and those who did not. The LIS at study entry was also not different when the patients were stratified into direct or indirect insult groups (Figure 1). Similarly, the Pa_O2/FI_O2 ratio was no different between those patients who developed ARDS and those who did not in direct and indirect insult groups (Figure 1).

View larger version (31K): [in this window] [in a new window]   Figure 1.   Boxplot showing median and interquartile ranges of the lung injury score, and PaO2/FIO2 ratio at study entry Day 0 (D0) and then daily until Day 3 (D3). The patients were stratified into direct and indirect lung insults. Mann-Whitney U tests were used to compare data between those who developed the acute respiratory distress syndrome (ARDS) and those who did not at the same time period; *p < 0.05, #p  0.01, +p  0.001. The lung injury score and PaO2/FIO2 ratio were no different between those patients who developed ARDS and those who did not on D0. However, over the next 3 d patients with ARDS developed more severe lung dysfunction.

The overall mortality was 28% (15 of 54), comprising nine and six patients with direct and indirect lung insults, respectively. Although mortality was no different between those patients developing ARDS and those who did not (p = 0.53), ventilator-free days were fewer in patients with ARDS (9 d [0-16] versus 19 d [10-24], p = 0.004). However, this was not significant when subdivided into indirect and direct lung insults (indirect lung insult: 21.0 [11.0-23.5] versus 15.0 [8.0-16.0], p = 0.11). Does not develop ARDS versus Develops ARDS. (Direct lung insult: 15.0 [1.5-24.0] versus 0.0 [0.0-15.0], p = 0.10).

Plasma SP-A and SP-B at Study Entry

When the patients were stratified according to their insult, both SP-A and SP-B were higher in the direct insult group (p = 0.02 and p < 0.001, respectively). Although there was no difference in plasma SP-A between those patients who developed ARDS and those who did not, plasma SP-B was significantly greater in the former group with a direct insult (Figure 2). Receiver operating characteristic analysis was performed on the LIS, plasma SP-A, and SP-B at study entry to examine their clinical utility as predictors of ARDS (Table 2). Whereas neither the AUC of the initial LIS nor plasma SP-A was different to the line of no information (AUC = 0.5), the AUC for plasma SP-B was significantly different (p = 0.0001 for all patients and p = 0.004 for patients with a direct lung insult). In patients with direct lung insult, an SP-B level of 4,994 ng/ml had a specificity of 78% (95% confidence interval: 40 to 97%), a sensitivity of 85% (55 to 98%), a positive predictive value of 85% (55 to 98%), and a negative predictive value of 78% (40 to 97%) (Figure 3).

View larger version (34K): [in this window] [in a new window]   Figure 2.   Boxplot of plasma surfactant protein-A (SP-A) and SP-B at study entry Day 0 (D0) and then daily until Day 3 (D3). The patients were stratified into direct and indirect lung insults. Mann-Whitney U tests were used to compare between patients who did and did not develop ARDS; *p < 0.05, #p  0.01, +p  0.001. Plasma SP-A was no different between those patients who developed ARDS and those who did not, and it was unchanged over the study period. At study entry plasma SP-B was greater in those patients with a direct lung insult who developed ARDS than in those who did not. In patients with an indirect lung insult who developed ARDS, plasma SP-B was no different at study entry, but gradually increased, and was similar to plasma SP-B in patients with a direct lung insult by D3 (p = 0.24).

View larger version (18K): [in this window] [in a new window]   Figure 3.   Receiver-operator characteristic curve for plasma surfactant protein-B (SP-B) at study entry. The optimal cutoff in patients with a direct lung insult was a SP-B concentration of 4,994 ng/ml, yielding a sensitivity of 85% and a specificity of 78%.

Ten of the 54 patients (five with a direct lung insult and five with an indirect lung insult) fulfilled the 1994 American-European Consensus Conference (AECC) definition of ARDS (14) at study entry. When these patients were excluded neither the LIS nor the plasma SP-A at Day 0 predicted the development of ARDS. However, plasma SP-B remained greater in patients with a direct lung insult (8,350 [4,567 to 17,001] versus 2,753 [2,434 to 2,487] pg/ml, p = 0.01; Develops ARDS versus Does not develop ARDS). Similarly, the AUC for the ROC analysis on Day 0 was significant only for plasma SP-B, both for all patients (AUC = 0.78 [0.64 to 0.91], p = 0.001), and for patients with a direct lung insult (AUC = 0.90 [0.77 to 1.03], p = 0.002).

Plasma SP-A and SP-B: Days 1-3

Plasma SP-B remained elevated in those patients who developed ARDS as a result of direct lung insult (Figure 2). However, in those who developed ARDS as a result of an indirect lung insult, plasma SP-B increased from study entry, and by Day 3 plasma SP-B was similar in both groups (p = 0.24). Plasma SP-A was unchanged over time in both groups of patients with ARDS.

Correlations

Plasma SP-A on Day 0 did not correlate with hospital mortality or ventilator-free days; however, a weak correlation was found between Day 0 plasma SP-B and ventilator-free days (r_s = 0.28, p = 0.04). No correlation was found between plasma SP-B and hospital mortality, and no correlation was found between plasma creatinine (152 [106 to 234] µmol/L) and plasma SP-B (r_s = 0.094, p = 0.52) or plasma SP-A (r_s = 0.085, p = 0.49).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Consistent with our previous reports (5, 6), we found elevated levels of circulating SP-A and SP-B in a broad range of patients both at-risk, and with ARDS. In the current study we identified a practical threshold for plasma SP-B levels that could be clinically useful in predicting the development of ARDS, particularly in patients suffering a direct lung insult.

We diagnosed ARDS using the LIS with bilateral infiltrates on chest radiograph. Although the 1994 AECC discarded the LIS and used instead the Pa_O2/FI_O2 ratio with bilateral infiltrates to define ARDS (14), the 1998 AECC has suggested a new stratification (the GOCA score), which includes a number of components included in the LIS (15). Both the LIS and the AECC definitions have been used to classify patients in recent trials targeted at reducing ventilator-induced lung injury in ARDS (16), and Meade and coworkers (19) have compared both definitions from a cohort of patients who had participated in a trial comparing two ventilation strategies. They found moderate agreement between both definitions, with some patients achieving ARDS criteria by one definition and not the other, and concluded that, in general, the two definitions could be used interchangeably. Further, similar patients are identified when these definitions are applied to defined at-risk patients (20). However, since the AECC definition relies on the Pa_O2/ FI_O2 ratio but fails to define the level of PEEP when this measurement is performed, we used the LIS to define ARDS.

Consistent with Meade and coworkers (19), ten of our 54 patients had ARDS at study entry according to the AECC definition. However, both with and without these patients in our cohort, plasma SP-B was significantly higher in patients who went on to develop ARDS when clinical criteria, represented by the LIS, were no different. Because an entry criterion for our study was the need for respiratory assistance, most of our patients already had some lung damage, and this was reflected in their elevated entry LIS, with many (41%) progressing to develop ARDS. This may be seen as a strength of the current study since few patients with a risk factor for ARDS are admitted to many intensive care units with normal lung function. Further, those patients who developed ARDS had a sustained elevation in LIS, whereas those patients who did not reach these criteria had a rapid decrease in their LIS over the next 3 d, and a greater number of ventilator-free days.

The Pa_O2/FI_O2 ratio of our patients at Day 0 was low, consistent with their need for respiratory assistance; however, it was not predictive of ARDS. The Pa_O2/FI_O2 ratio is influenced by many factors in addition to lung injury. For example, patients with unilateral infiltrates on chest radiograph such as lobar pneumonia or focal aspiration do not fulfill ARDS definitions despite severe hypoxemia since they do not fulfill the radiographic criteria. Consistent with our LIS data, and partly because the Pa_O2/FI_O2 ratio contributes to the LIS, patients who developed ARDS gradually developed worse oxygenation than those who did not. However, the LIS appeared to separate more than the Pa_O2/FI_O2 ratio by Day 3. This likely reflects the additional contribution of PEEP, respiratory system compliance, and chest radiograph score to the LIS, whereas the Pa_O2/FI_O2 ratio is strongly dependent upon clinical management.

The proportion of at-risk patients developing ARDS varies depending upon the insult, but tends to be highest with sepsis and aspiration of gastric contents, both of which also tend to lead to the early development of ARDS (21, 22). If pneumonia is included as part of sepsis, 61% of our patients had a high-risk factor. Further, Sharkey and colleagues (23) and Connelly and colleagues (24) had ARDS rates of 38 and 36%, respectively, suggesting that our patients represent a similar cohort; however, neither study documented the respiratory state of their patients at study entry.

Irrespective of the ARDS definition used, the aim is to define a clinical syndrome with a minimum severity of acute respiratory dysfunction, and the relationship of this to the severity of DAD, and to the process of repair and fibrosis will be influenced by many factors. The rationale for our study was that DAD, with an increase in alveolocapillary permeability, precedes the clinical diagnosis of ARDS. Indeed, disease pathogenesis must always precede clinical manifestations. Although Greene and coworkers (7) did not examine plasma SP-B levels, consistent with this rationale they found that plasma SP-A was weakly predictive of ARDS in septic patients. However, in a separate cohort, serum SP-A was not elevated in the at-risk patients, but was in patients with ARDS, reaching its maximum on Day 3 (8). Because the major barrier to protein leakage across the alveolocapillary barrier is the epithelium, and because SP-B is considerably smaller than SP-A, it is not surprising that (1) plasma SP-B, but not SP-A, was predictive of ARDS, and that (2) this was apparent in patients with a direct lung insult where an earlier epithelial injury would be expected.

There is growing interest in the pathophysiologic differences between direct and indirect causes of ARDS. Many of the indirect or extrapulmonary causes of ARDS result in reduced chest wall and abdominal compliance with respiratory mechanics suggestive of alveolar collapse (25). Consequently, clinical markers of ARDS severity, such as respiratory system compliance and the Pa_O2/FI_O2 ratio, which are influenced by abdominal and chest wall pathology, may not accurately reflect the severity of DAD in patients with an indirect insult. Consistent with this, we found an earlier increase in plasma SP-B in patients with direct lung injury. Further, plasma SP-B levels in patients who developed ARDS because of an indirect insult rose over 3 d to the same level as those initially found in patients who developed ARDS caused by a direct insult, despite similar LIS. This is consistent with a different pattern of DAD pathogenesis and altered epithelial permeability, as might be predicted with an indirect, as opposed to a direct, insult. Although we assessed the severity of DAD in patients with ARDS using the LIS, other simple clinical indices such as the Pa_O2/FI_O2 ratio or alveolar-to-arterial O₂ difference will also be influenced by abnormal chest wall and abdominal compliance. Indeed, none of these indices accurately reflect the extent or pathogenesis of DAD (26).

The surfactant proteins have many features to recommend them as biologic markers in ARDS. It is uncertain precisely how they breach the alveolocapillary barrier, but consistent with a bidirectional fluid flux across the alveolocapillary membrane, we have estimated that the healthy lung maintains an ELF:plasma gradient of ~ 7,000:1 and ~ 1,500:1 for SP-A and SP-B, respectively. However, when the alveolocapillary barrier is injured, these lung-specific proteins are no longer effectively partitioned and increased amounts leak into the bloodstream. In addition, their brief systemic half-life suggests that their circulating levels acutely reflect changes in lung permeability (27). Consequently, circulating surfactant proteins allow a simple, potentially clinically applicable measurement of this unique partitioning function of the lung.

Although many biologic markers are elevated in at-risk patients, few have proven predictive for the development of ARDS. Donnelly and coworkers (28) reported that plasma soluble L-selectin levels were significantly lower in at-risk patients who subsequently developed ARDS. Similarly, Parsons and coworkers (29) examined plasma IL-ra and IL-10 levels in 77 at-risk patients, and although both cytokines were elevated, neither was predictive for ARDS. This likely reflects the lack of specificity of markers that rely on a systemic inflammatory response. In contrast, serum ferritin levels do appear to be predictive (23, 24). Clinically useful cutoff levels have been identified in a broad range of at-risk patients, including those with multiple trauma. Although the mechanism for the elevation in serum ferritin is uncertain, it must reflect a systemic response to a risk factor, which may prove to reduce its specificity.

Another variable in the predictive power of a given biomarker is the precise question investigated. For example, whereas Ruben and coworkers (30) found that plasma von Willebrand factor antigen, an endothelial marker, was predictive of acute lung injury in 45 patients with nonpulmonary sepsis and an entry LIS of zero, Moss and associates (31) found it was not when applied to a broader range of at-risk patients. This discrepancy may reflect differences in the extent and timing of endothelial activation in different groups of at-risk patients.

There are two main questions when examining a biologic marker for prediction of ARDS: (1) how early after an insult can biologic changes be detected, and (2) of greater practical importance to the intensivist is which ICU patients requiring ventilatory assistance will develop ARDS? This study addresses the latter. From a pragmatic clinical perspective, a useful biologic marker must add information that is not apparent from routine examination and investigation. LIS was not predictive of ARDS. Plasma SP-B was elevated on study entry, indicating that significant alveolocapillary damage was present at this time. We do not know how early elevated plasma SP-B precedes respiratory dysfunction. However, the capacity of plasma SP-B to predict ARDS in these already compromised patients emphasizes its merit since arguably a more distinct cutoff would have been found if we had studied these patients earlier in the pathogenesis of their lung injury. A similar argument can also be made had we included at-risk patients who did not subsequently require intensive care admission and support. Nevertheless, the current study design does examine the clinically important issue of ARDS prediction in critically ill patients.

Pharmacologic modification of ARDS may soon be possible with our growing insight into the pathogenesis and pathophysiology of DAD. Specific biomarkers of epithelial and endothelial dysfunction, inflammation, and repair may allow early prediction of ARDS, and the ability to monitor these processes during treatment. However, further research is needed. This includes larger studies examining the surfactant proteins in at-risk patients with less severe pulmonary dysfunction than our cohort, and studies comparing and combining biomarkers, and with adequate power to analyze specific at-risk groups.