The primary objectives of this study was to test the correlation between pulmonary oedema assessed using the single indicator transpulmonary thermodilution EVLWI measurement and a validated chest radiograph score in patients with ALI/ARDS as well as to test their predictive value for mortality. There was only a moderate positive correlation between the chest radiograph findings and the EVLWI measurement. This correlation was lower, but similar to those of prior studies done more than 20 years ago in which the correlation ranged from r
]. As a dichotomous test, a chest radiograph score >85 was sensitive in determining whether EVLW is present, but a score >85 was not specific for EVLW ≥16 ml/kg.
Using the single-indicator transpulmonary thermodilution method to measure EVLWI is attractive for many reasons. First, this method generates a quantitative measure of EVLW, in contrast to the chest radiograph, which must be interpreted qualitatively by clinicians and is susceptible to interobserver disagreement. We tried to reduce interobserver variability by using an objective validated chest radiograph scoring system [33
]. Once ALI/ARDS is established, current practice relies on clinicians accurately interpreting successive chest radiographs to determine whether pulmonary oedema is accumulating or resolving. Although chest radiographs are relatively inexpensive, readily available, and noninvasive, quantitative measures are clinically more useful than qualitative measures. For example, specific cutoffs in a quantitative measure such as EVLWI could be used to guide patient management in a fluid management protocol. Second, the chest radiograph is insensitive to small changes in the quantity of EVLW present [36
]. In contrast, small changes (10-20%) in EVLW can be detected using the single indicator method [37
]. Detecting small changes in EVLWI may allow for earlier diagnosis of ALI/ARDS and more precise monitoring of responses to therapeutic measures. Finally, lung inflation affects the appearance of the chest radiograph; likewise, atelectasis may be difficult to differentiate from airspace opacities consistent with alveolar oedema [38
]. However, these potential benefits of the transpulmonary thermodilution method must be weighed against the potential complications of this invasive method, including the risk of bleeding at the site of vascular access, arterial injury, and thrombosis.
The change in daily chest radiograph scores was not predictive of time to death. In contrast, EVLWI changes over consecutive ICU days predicted mortality. This builds upon our previous data that baseline EVLWI
16 ml/kg predicts mortality in ALI [32
]. Other studies have reported that EVLW may be an independent predictor of mortality for ICU patients [4
]. One study reported a mortality rate of 65% when EVLW was >15 ml/kg [4
], and another found that EVLWI >16 ml/kg predicted mortality with 100% specificity and 86% sensitivity [3
]. In addition, EVLWI may have predictive value during the entire clinical course because maximum EVLWI is a predictor of mortality [3
]. This is consistent with our finding that the change in EVLWI measurements over the ICU course was predictive of mortality. Taken together, these observations indicate that EVLWI measurements may provide a quantitative assessment of lung injury severity, which could be used to guide therapy in ALI/ARDS. Consistent with this, prior studies have reported that EVLWI measurements can be successfully used to monitor response to β agonist therapy [39
]. The next step would be to design a clinical trial in which EVLWI measurements are used to guide fluid and diuretic therapy, and within this same trial, determine if decreases in EVLW are associated with a decrease in mortality.
The threshold of 16 ml/kg for the EVLWI has been used to identify ARDS patients with a higher mortality. This level indicates a high amount of pulmonary oedema, although levels between 10 and 15 ml/kg represent progressive interstitial and alveolar oedema [41
]. Some investigators have concluded that as many as 50% of patients with a clinical diagnosis of ARDS have a normal EVLWI [42
], but more prospective data are needed to assess this issue.
There was a trend towards a lower baseline PaO2
ratio in nonsurvivors, and in logistic regression baseline PaO2
ratio predicted ventilator free days and mortality. In addition, the change in PaO2
ratio over consecutive ICU days predicted time to death. Although not consistent in all studies, baseline PaO2
ratio has been reported in some studies to be lower in nonsurvivors [43
] and predicts mortality in univariate analyses [43
]. In addition, in one large cohort study, PaO2
ratio was an independent predictor of mortality [44
]. However, these studies did not test the change in PaO2
ratio over consecutive ICU days as a predictor of mortality. A limitation of PaO2
ratio as a predictor of outcome is the fact that it can be modified independently by adjustment in ventilatory settings, a limitation to which measurement of EVLWI is not susceptible.
The baseline chest radiograph score differed between survivors and non-survivors, but in contrast to EVLWI and PaO2
, did not predict mortality. The change in daily chest radiograph scores did not predict time to death although it did predict ventilator-free days. Most prior studies of chest radiograph findings in patients with pulmonary oedema examined the ability of the chest radiograph to predict the amount of EVLW present, not mortality [12
]. Our chest radiograph scoring system is easy to apply clinically, has been validated [33
], and has a high sensitivity for the detection of EVLW.
Measurement of the EVLWI using the transpulmonary thermodilution method may serve two important purposes. First, measuring EVLWI using this method may be useful to confirm the presence of an increase in pulmonary oedema. In addition, by using the transpulmonary thermodilution method, EVLWI is measured quantitatively. Trending the EVLWI measurements over consecutive ICU days predicts mortality and may be of value for guiding therapy.
Our study has some limitations. First, the total number of deaths (18/59 patients) limited our ability to perform a multivariable regression analysis controlling for potentially confounding conditions that may affect the risk of death. However, because we collected data on multiple days for each patient, we did have sufficient power to detect associations between the chest radiograph score and EVLWI. More importantly, we could use Cox regression models to determine whether the change in each of these predictors (chest radiograph score, EVLWI, and PaO2
) over time predicted time to death. Second, when scoring the chest radiographs, we only considered the proportion of the airspace that was affected, and not the density of the alveolar oedema. Certain characteristic signs, such as pulmonary congestion and vascular redistribution, are associated with small increases in EVLW [8
]. As the quantity of EVLW increases, densities occupy a greater proportion of the total airspace, and as EVLW further increases, the density of the airspace opacities also worsens [8
]. If we had accounted for the density in addition to the proportion of airspace affected, the correlation between the chest radiograph score and EVLWI might have been stronger, and chest radiograph score may have been predictive of mortality. Third, many of the chest radiographs were scored at the maximum chest radiograph score. It would have been ideal to have chest radiograph scores distributed more evenly across the entire scoring range to provide more variability to determine how these scores match with the wide range of EVLWI measurements. Fourth, we did not use the PiCCO device to estimate lung vascular permeability, which other investigators have used [48
]. Fifth, we did not quantify pleural effusions by ultrasound or CT [49
], and sixth, we did not quantify fluid balance in these patients [50