Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Open Access

Peer-reviewed

Research Article

Association between Different Indexations of Extravascular Lung Water (EVLW) and PaO2/FiO2: A Two-Center Study in 231 Patients

  • Wolfgang Huber ,

    wolfgang.huber@lrz.tum.de

    Affiliation II. Medizinische Klinik und Poliklinik, Klinikum rechts der Isar der Technischen Universität München, München, Germany

    ⨯
  • Josef Höllthaler,

    Affiliation II. Medizinische Klinik und Poliklinik, Klinikum rechts der Isar der Technischen Universität München, München, Germany

    ⨯
  • Tibor Schuster,

    Affiliation Institut für Medizinische Epidemiologie und Statistik, Klinikum rechts der Isar der Technischen Universität München, München, Germany

    ⨯
  • Andreas Umgelter,

    Affiliation II. Medizinische Klinik und Poliklinik, Klinikum rechts der Isar der Technischen Universität München, München, Germany

    ⨯
  • Michael Franzen,

    Affiliation II. Medizinische Klinik und Poliklinik, Klinikum rechts der Isar der Technischen Universität München, München, Germany

    ⨯
  • Bernd Saugel,

    Affiliation II. Medizinische Klinik und Poliklinik, Klinikum rechts der Isar der Technischen Universität München, München, Germany

    ⨯
  • Colin Cordemans,

    Affiliation Department of Intensive Care, Ziekenhuis Netwerk Antwerpen, Antwerpen, Belgium

    ⨯
  • Roland M. Schmid,

    Affiliation II. Medizinische Klinik und Poliklinik, Klinikum rechts der Isar der Technischen Universität München, München, Germany

    ⨯
  • Manu L. N. G. Malbrain

    Affiliation Department of Intensive Care, Ziekenhuis Netwerk Antwerpen, Antwerpen, Belgium

    ⨯

Abstract

Background

Variability of body weight (BW) and height calls for indexation of volumetric hemodynamic parameters. Extravascular lung water (EVLW) has formerly been indexed to actual BW (BWact) termed EVLW-index (EVLWI). In overweight patients indexation to BWact might inappropriately lower indexed EVLWIact. Several studies suggest indexation of EVLWI to predicted BW (EVLWIpred). However, data regarding association of EVLWIact and EVLWpred to mortality and PaO2/FiO2 are inconsistent. Two recent studies based on biometric database-analyses suggest indexation of EVLWI to height (EVLWIheight). Therefore, our study compared the association of un-indexed EVLW, EVLWIheight, EVLWpred and EVLWIact to PaO2/FiO2 and Oxygenation index (OI = mean airway pressure*FiO2*/PaO2).

Methods

A total of 2119 triplicate transpulmonary thermodilutions (TPTDs; PiCCO; Pulsion Medical-Systems, Germany) were performed in 50 patients from the evaluation, and 181 patients from the validation groups. Correlations of EVLW and EVLWI to PaO2/FiO2, OI and ROC-AUC-analyses regarding PaO2/FiO2<200 mmHg (primary endpoint) and OI>10 were performed.

Results

In the evaluation group, un-indexed EVLW (AUC 0.758; 95%-CI: 0.637-0.880) and EVLWIheight (AUC 0.746; 95%-CI: 0.622-0.869) provided the largest ROC-AUCs regarding PaO2/FiO2<200 mmHg. The AUC for EVLWIpred was smaller (0.713). EVLWIact provided the smallest AUC (0.685). This was confirmed in the validation group: EVLWIheight provided the largest AUC (0.735), EVLWIact (0.710) the smallest. In the merged data-pool, AUC was significantly greater for EVLWIheight (0.729; 95%-CI: 0.674–0.784) compared to all other indexations including EVLWIact (ROC-AUC 0.683, p = 0.007) and EVLWIpred (ROC-AUC 0.707, p = 0.015). The association of EVLW(I) was even stronger to OI compared to PaO2/FiO2. In the merged data-pool, EVLWIheight provided the largest AUC regarding “OI>10” (0.778; 95%-CI: 0.713–0.842) compared to 0.739 (95%-CI: 0.669–0.810) for EVLWIact and 0.756 (95%-CI: 0.688–0.824) for EVLWIpred.

Conclusions

Indexation of EVLW to height (EVLWIheight) improves the association of EVLW(I) to PaO2/FiO2 and OI compared to all other indexations including EVLWIpred and EVLWIact. Also considering two recent biometric database analyses, EVLWI should be indexed to height.

Citation: Huber W, Höllthaler J, Schuster T, Umgelter A, Franzen M, Saugel B, et al. (2014) Association between Different Indexations of Extravascular Lung Water (EVLW) and PaO2/FiO2: A Two-Center Study in 231 Patients. PLoS ONE 9(8): e103854. https://doi.org/10.1371/journal.pone.0103854

Editor: Jörn Karhausen, Duke University Medical Center, United States of America

Received: March 25, 2014; Accepted: July 3, 2014; Published: August 5, 2014

Copyright: © 2014 Huber et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: The authors confirm that, for approved reasons, some access restrictions apply to the data underlying the findings. Due to ethical and legal restrictions, confidential data are available upon request. To receive anonymized data readers are welcome to contact the corresponding author (Wolfgang Huber) for the Munich collective and Manu Malbrain for the Antwerp data. Prof. Dr. Wolfgang Huber, II. Medizinische Klinik und Poliklinik, Klinikum rechts der Isar der Technischen Universität München, Ismaninger Strasse 22, D-81675 München, Germany. Tel.: 0049-89-4140-2265. Fax.: 0049-89-4140-4808. E-mail: wolfgang.huber@lrz.tum.de. Prof. Manu L.N.G. Malbrain, Department of Intensive Care, Ziekenhuis Netwerk Antwerpen, ZNA Stuivenberg, Lange Beeldekensstraat 267, B-2060 Antwerpen 6, Belgium.

Funding: The authors have no support or funding to report.

Competing interests: Wolfgang Huber, Bernd Saugel and Manu LNG Malbrain received lecture fees and are members of the Medical Advisory Board of Pulsion Medical Systems, Feldkirchen, Germany. All other authors have no conflict of interest. Hereby the authors confirm that this does not alter their adherence to PLOS ONE policies on sharing data and materials.

Introduction

Extravascular lung water (EVLW) is a measure of the interstitial, alveolar and lymphatic fluid content of the lungs. EVLW and its indexation to body weight (EVLWI) became routinely available after the introduction of single-indicator trans-pulmonary thermodilution (TPTD) [1]–[5]. A number of animal and clinical studies demonstrated an association of EVLW(I) to mortality and to parameters of pulmonary function such as PaO2/FiO2 [5]–[19]. Variability of body weight (BW) and height strongly calls for biometric adjustment and indexation. Ideally, appropriate indexation should be based on a limited number of routinely available biometric data, and it should result in consistent normal values for patients with different height, weight, age, gender and race [20]–[21]. Originally actual body weight (BWact) was used for indexation of EVLW. However, triggered by a rapidly increasing number of obese patients [22], [ 23], the question arose as to which weight to choose for EVLW indexation since indexation to BWact might inappropriately decrease EVLWIact in obese patients. Based on a better correlation to mortality, a number of studies have suggested indexation of EVLW to predicted BW (EVLWIpred) (see Table 1) [12: 14]. However, data are inconsistent regarding other endpoints: e.g. if available data provide worse [12], similar [14], [ 16] or slightly better correlation of EVLWIpred to PaO2/FiO2 than EVLWIact [13]. In one of the most recent studies [15], “Chew et al. found that EVLW indexed to absolute body weight resulted in a stronger association with outcome” including mortality compared to EVLWIpred [24]. Despite the overall strong predictive capacity of EVLWI in this and other recent trials [12]–[19], these inconsistencies demonstrate the need to optimize indexation of EVLW. Therefore, we recently analyzed a prospectively maintained database regarding the association of EVLW to biometric data [18]. This study demonstrated that height was the only biometric parameter independently associated to EVLW. These data were recently confirmed by Wolf et al., using a similar approach in a surgical group [19]. Despite these conclusive data, both studies did not investigate, if indexation of EVLW to height (EVLWIheight) provides better association to pulmonary function and outcome. Furthermore, all these trials were mono-centric.

thumbnail
Download:
Table 1. Indexations of extravascular lung water (index) EVLW/EVLWI.

https://doi.org/10.1371/journal.pone.0103854.t001

Therefore, this two-center study compared the association of PaO2/FiO2 (and other outcome markers) to EVLW (un-indexed), EVLWIact, EVLWIpred, EVLWIheight and EVLWI indexed to other biometric indices. To provide sufficient balance of biometric data, we investigated a group with a representative distribution of body mass index (BMI) [22], as well as an unselected second group from a second center.

Materials and Methods

Munich-evaluation-group

The institutional ethics committee approved the study (Ethikkommission Technische Universität München; Fakultät für Medizin; No. 3049/11). Patients on mechanical ventilation monitored using TPTD regardless of the study were included in the prospectively maintained database. The need for informed consent was waived due to the non-interventional design of the study. The patients included in this study completely distinct to the group previously analysed regarding the association of EVLW to biometric data [18].

To provide a representative distribution regarding bodyweight [22], we included 15 consecutive patients with BMI≥30 kg/m2, 15 consecutive patients with 25≤BMI<30 kg/m2 and 20 consecutive patients with a normal BMI (<25 kg/m2) irrespective of fulfilling the criteria of acute respiratory distress syndrome (ARDS). Conscious patients were asked for actual biometric data. In unconscious patients body weight and height were extracted from the patients records. In case of doubt height was verified using a flexible tape measure in the supine position.

A 5-F thermistor-tipped femoral arterial line (PV2025L20, Pulsiocath, Pulsion Medical Systems, Munich, Germany) connected to the PiCCO monitor device (PiCCO-Plus; Pulsion Medical Systems) was used for TPTD measurements. The mean EVLW was measured based on TPTD performed in triplicate with 15 ml cold saline 0.9%.

Antwerp-validation-group

Retrospective analysis of data from a prospectively developed independent cohort was performed for the first 7 days of ICU admission. Data of 181 critically ill patients requiring mechanical ventilation and TPTD-hemodynamic monitoring treated in two ICU's in ZNA Campus Stuivenberg, Antwerp, Belgium were collected prospectively. Ethics approval had been obtained and due to the retrospective analysis and non-intervention-nature of the study the need for informed consent was waived (project number EC 3765; Commisie voor Medische Ethiek, Ziekenhuisnetwerk Antwerpen 2020). Parts of the data not related to EVLW-indexation have already been published in Annals of Intensive Care [25], [ 26].

The measuring technique was identical to the Munich-evaluation-group, with the only difference being that three 20 ml boluses of cooled saline were used for TPTD.

Mean length of the ICU-stay in the Munich group was 27.2±21.4 days with a range of 3 to 120 days. In the Antwerp group the mean ICU stay was 25.9±41.7 days with a range of 1to 429 days.

Endpoints and Statistics

thumbnail
Download:
Table 2. Patients' characteristics.

https://doi.org/10.1371/journal.pone.0103854.t002

  1. Weight and weight-correction formula based indexations. To compare the distribution of normal (EVLWI<7 ml/kg), slightly elevated (7≤EVLWI<10 ml/kg) and markedly elevated EVLWI (EVLWI≥10 ml/kg) in dependency of the indexation, un-indexed EVLW was indexed according to actual (EVLWIact), predicted (EVLWIpred), ideal (EVLWIid) and adjusted BW (EVLWIadj) using the formulas mentioned in Table 1. The cut-offs of 10 ml/kg and 14 ml/kg have been demonstrated to be associated with ARDS [15] and mortality [5].
    1. –Intra-group comparisons: For the weight-related indexations, intragroup comparisons of weight-correction based EVLWIpred, EVLWIid and EVLWIadj to EVLWIact were performed for all patients and the subgroups with BMI<25 kg/m2, 25≤BMI<30 kg/m2 and BMI≥30 kg/m2 (Wilcoxon-test for paired samples; Table 2).
    2. –Inter-group comparisons: Furthermore, we compared EVLWI according to all investigated indexations (also including EVLWIheight, EVLWIBMI, EVLWIBSA and EVLWITLC) between patients with BMI<25 kg/m2 and patients with BMI≥30 kg/m2 (Wilcoxon-test for unpaired samples; Table 2).
  2. Association of EVLW(I) to PaO2/FiO2 and oxygenation index (OI =  mean airway pressure * FiO2 * PaO2−1). For appropriate analysis of multiple serial measurements in 231 patients from the two groups several statistical analyses were performed:
    1. 2a). Prediction of critical thresholds of PaO2/FiO2 and oxygenation index: The clinically relevant prediction of critical thresholds of “PaO2/FiO2<200 mmHg” (primary endpoint) and “OI>10” by EVLW(I) was investigated using receiver operating characteristics area under the curve (ROC-AUC) analyses of all measurements.
    2. 2b). Inter-individual (“between-subject”) correlations: Furthermore, correlations of EVLW and differently indexed EVLWI to PaO2/FiO2 and oxygenation index (OI =  mean airway pressure * FiO2 * PaO2−1) were calculated. Since multiple serial measurements within 241 different patients were available, we analysed inter- and intra-individual correlations.
    To correct for different numbers of measurements for each patient, the means of EVLW(I), PaO2/FiO2 and OI were calculated for each individual patient (“one point per patient”). Subsequently the correlations between EVLW(I) and PaO2/FiO2 and between EVLWI and OI were calculated.
    1. 2c). Intra-individual (“within subject”) correlations: The above-mentioned “one point per patient” analyses reflect the inter-individual association of EVLW(I) to PaO2/FiO2 and OI. However, in cases of multiple serial measurements within different patients, between-subject heterogeneity may obscure correlations on an individual level (within subject correlation) which might be even more interesting than the inter-individual association. The effect of the confounder (between-subject heterogeneity) can be removed by calculating “partial” correlation between EVLW(I) and PaO2/FiO2 (or OI) adjusting for heterogeneity of different patients (individual patient number/identifier as the adjustment factor).
  3. Mortality analysis. Better prediction of mortality by EVLWI according to any indexation might be related to the direct association of the indexation to mortality. To overcome this problem, multiple binary logistic regression analysis regarding mortality included the first and last values of unindexed EVLW as well as BWact, height, gender and acute physiology and chronic health evaluation (version 2, APACHE-II).

All analyses were performed separately in both groups and in the merged data, with the only exceptions being intergroup-comparisons and mortality analyses which were restricted to the BMI-representative Munich-evaluation-group. Results of merged data were considered superior to those derived from sub-groups. No correction of p-values was applied to adjust for multiple testing. However, results of all statistical tests being conducted were thoroughly reported so that an informal adjustment of p-values can be performed while reviewing the data [27].

All statistical tests were conducted 2-sided and a p-value <0.05 was considered to indicate statistical significance. The software used was IBM SPSS statistics, version 20.

Results

Patients' characteristics

A total of 2119 TPTDs (each with triplicate TPTD) were performed in 231 patients from both groups.

Table 2 summarizes the patients' characteristics of both groups.

  1. 1. Weight and weight-correction formula based indexations. Comparison of EVLWIact, EVLWIpred, EVLWIid and EVLWIadj: Table 2 demonstrates that in the Munich-evaluation-group mean values of EVLWIpred, EVLWIadj and EVLWIid were significantly higher than EVLWIact in the subgroups of patients with BMI≥30 kg/m2 and with 25≤BMI<30 kg/m2 as well as for the totality of patients (intra-BMI-group-comparison). By contrast, EVLWIpred and EVLWIadj were significantly lower than EVLWIact in the subgroup of patients with a normal BMI.

Similarly, in the Antwerp-validation-group mean values of EVLWIpred, EVLWIadj and EVLWIid were higher than mean the EVLWIact.

Impact of indexation according to different weight correction-formulas for the classification of EVLWI: Distribution of EVLWI-values classified as normal (EVLWI<7 ml/kg), moderately elevated (7 ml/kg≤EVLWI<10 ml/kg) and markedly elevated EVLWI (EVLWI≥10 ml/kg) significantly varied among the patients with a BMI≥30 kg/m2 as well as in the total patient groups depending on the weight used for indexation of EVLWI (Fig. 1). For example in patients with a BMI≥30 kg/m2, 51% (133/263) of EVLWIact measurements were within the normal range (EVLWI<7 ml/kg). By contrast only 16% (43/263; p<0.001 vs. EVLWIact), 14%, (38/263; p<0.001) and 30% of the measurements (79/263; p<0.001) were within the normal range if EVLWI was indexed according to predicted, ideal and adjusted BW, respectively. In addition to the different distributions of EVLWI classifications, different indexations obviously had an impact on the coefficient of variation (COV), in particular in patients with a BMI≥30 kg/m2 (Table 2). COV amongst these patients was markedly lower for EVLWIact (0.41) than for EVLWIheight (0.57) or un-indexed EVLW (0.60). Inter-BMI-group-comparison demonstrated significantly lower values in patients with a BMI≥30 kg/m2 for EVLWIact, EVLWIBMI, EVLWIBSA and EVLWIadj, whereas there was no inter-group difference for EVLW, EVLWIpred, EVLWIid, EVLWIheight and EVLWITLC. In conclusion, significant inter-BMI-group-differences were found only for EVLWIact and indexations including BWact in their formulas (BWadj, BMI and BSA) (Table 2).

thumbnail
Download:
Figure 1. Distribution of measurements classified as “normal” (EVLWI<7 mL/kg), “moderately elevated” (7≤EVLWI<10 mL/kg) and “markedly elevated” (EVLWI≥10 mL/kg) depending on the indexation of EVLWI according to weight-based indexations to BWact, BWpred, BWid and BWadj.

Numbers in the columns indicate the percentage of measurements within this classification (*p<0.05 and ** p<0.001 vs. percentage of normal measurements of EVLWIact<7 mL/kg within the same BMI-category).

https://doi.org/10.1371/journal.pone.0103854.g001

  1. 2. Association of EVLW(I) to parameters of pulmonary function. 2a.) Prediction of “PaO2/FiO2<200 mmHg” (primary endpoint) and “OI>10”: As demonstrated in Table 3, in the Munich-evaluation-group the greatest ROC-AUCs regarding “PaO2/FiO2<200 mmHg” were found for un-indexed EVLW (ROC-AUC 0.758; 95%-CI: 0.637–0.880) and EVLWIheight (ROC-AUC 0.746; 95%-CI: 0.622–0.869). EVLWIact provided the lowest ROC-AUC (0.685, 95%-CI: 0.554–0.817)
thumbnail
Download:
Table 3. Comparison of receiver operating characteristics area under the curve (ROC-AUC) regarding „PaO2/FiO2<200“ and “Oxygenation-Index (OI)>10” depending on indexation of extravascular lung water (index) EVLW(I): ROC-AUCs for different indexations of EVLW(I) regarding „PaO2/FiO2<200“ (left side of the table) and “OI>10” (right side of the table).

https://doi.org/10.1371/journal.pone.0103854.t003

In general, these observations were confirmed in the Antwerp-validation-group: EVLWIheight had the highest predictive capability (ROC-AUC 0.735; 95%-CI: 0.674–0.796), whereas weight-indexed EVLWIact (ROC-AUC 0.710; 95%-CI: 0.648–0.773) and EVLWIBMI (ROC-AUC 0.704; 95%-CI: 0.641–0.767) provided the smallest ROC-AUCs.

Statistical analysis of the merged data of both groups demonstrated a number of significant differences between different indexations, summarized as follows:

  1. 1. The greatest ROC-AUC regarding “PaO2/FiO2 <200 mmHg” was found for EVLWIheight (ROC-AUC 0.729; 95%-CI: 0.674–0.784; primary endpoint)
  2. 2. The ROC-AUC was significantly greater for EVLWIheight compared to all other indexations including EVLWIact (ROC-AUC 0.683; 95%-CI: 0.626–0.741; p = 0.007) and EVLWIpred (ROC-AUC 0.707; 95%-CI: 0.650–0.763; p = 0.015). Only un-indexed EVLW (ROC-AUC 0.728; 95%-CI: 0.673–0.783) and EVLWIBSA (ROC-AUC 0.718; 95%-CI: 0.663–0.774; p = 0.137) were not significantly inferior compared to EVLWheight.

Regarding the prediction of the threshold “OI>10”, in both collectives as well as in the merged data, the largest ROC-AUCs were obtained for EVLWIheight and EVLW, with the lowest for EVLWIact (Table 4): 0.737 (0.589–0.885), 0.732 (0.583–0.881) and 0.669 (0.502–0.835) in the Munich-evaluation-group, 0.778 (0.713–0.842), 0.771 (0.705–0.836), and 0.739 (0.669–0.810) in the Antwerp-validation group and 0.762 (0.702–0.822), 0.756 (0.695–0.817) and 0.716 (0.649–0.782) for the merged data, respectively.

thumbnail
Download:
Table 4. Intra- and inter-individual correlations of extravascular lung water (index) EVLW(I) to PaO2/FiO2.

https://doi.org/10.1371/journal.pone.0103854.t004

For the merged data, the AUC was significantly higher for EVLWIheight (0.762) compared to EVLWIact (0.716; p = 0.017), EVLWIadj (0.739; p = 0.05) and EVLWIBMI (0.719; p = 0.019).

2b.), 2.c) Inter-individual and intra-individual correlations of EVLW(I)to PaO2/FiO2 and to oxygenation index: In general, the inter-individual association (represented by patients' means) of EVLW(I) to PaO2/FiO2 and OI was not as strong as the intra-individual association represented by partial correlations. Means of EVLW(I) and PaO2/FiO2 moderately correlated with r-values between −0.21 and −0.46. By contrast, intra-individual correlations represented by partial correlations provided r-values between −0.6 and −0.74 (Table 4).

Within the Munich-evaluation-group only the patients' means of EVLW, EVLWIheight and EVLWIadj significantly correlated with PaO2/FiO2 (r = −0.34; p = 0.017, r = −0.32, p = 0.026 and r = −0.29; p = 0.041, respectively). In contrast, the patients' mean EVLWI according to all other indexations including EVLWIpred and EVLWIact did not correlate to PaO2/FiO2. Overall, the lowest r-values were found for EVLWIact. The highest coefficients of correlations to PaO2/FiO2 were found for un-indexed EVLW and EVLWIheight (Table 4).

However, in the same group all partial correlations of EVLW(I) and PaO2/FiO2 were highly significant (p<0.001) and provided coefficients of partial correlation between −0.66 and −0.74.

Similar data were obtained for the correlation between OI and EVLW and its indexations (Table 3). Intra-individual partial correlations with r-values between 0.57 and 0.80 were more pronounced than inter-individual correlations represented by r-values between 0.31 and 0.48 for correlations of patients' means. All these correlations were significant. In both groups as well as in the merged data the highest coefficients of correlation to OI were found for EVLWIheight and EVLW, the lowest for EVLWIact (r-values for merged data 0.77, 0.78 and 0.69, respectively) (Table 4; partial correlation).

  1. 3. Mortality analysis. Univariable logistic regression analysis of the Munich-evaluation-group demonstrated a significant association between mortality with APACHE-II-Score (p = 0.005; 95%-CI: 1.063–1.422; β-coefficient of regression 0.207), but not with age (p = 0.262), height (p = 0.265) and BWact (p = 0.123). The strongest association to mortality was found for the last EVLW (p = 0.001; 95%-CI: 1.185–2.035; β-coefficient of regression 0.440 for increments in EVLW of 100 ml). The first EVLW was associated with mortality with borderline significance (p = 0.054; 95%-CI: 0.098–1.350; β-coefficient of regression 0.149).

Subsequently, a multivariable logistic regression analysis regarding mortality was performed. A model including APACHE-II, the first and the last EVLW provided high predictive capabilities regarding mortality (Nagelkerkes R2 = 0.697). First (p = 0.021) and last (p = 0.004) EVLW were independently associated to mortality. The APACHE-II score slightly failed to reach significance (p = 0.064), but markedly contributed to the R2-value of the total model (R2 = 0.628 without APACHE-II).

ROC-analysis (Figure 2) regarding mortality demonstrated high predictive capabilities of the model including APACHE-II as well as first and last EVLW (AUC 0.936; 95%-CI: 0.868–1.000; p<0.001), which provided a markedly larger ROC-AUC than each of the included single parameters. Among single parameters, the last EVLW (AUC 0.868; 95%-CI: 0.765–0.970; p<0.001) and APACHE-II (AUC 0.779; 95%-CI: 0.636–0.923; p = 0.002) provided high predictive capabilities compared to the first EVLW (AUC 0.603; 95%-CI: 0.424–0.782; p = 0.244).

thumbnail
Download:
Figure 2. ROC curve regarding the prediction of mortality provided by first EVLW, last EVLW, APACHE-II and regression model combining first EVLW, last EVLW and APACHE-II.

https://doi.org/10.1371/journal.pone.0103854.g002

Discussion

Data regarding EVLW-indexation are contradictory. Therefore, our study investigated the association between different indexations of EVLW(I) to PaO2/FiO2 and OI in two groups with mechanical ventilation and representative distribution of BMI. Our study demonstrated that

  1. indexation of EVLWI to BWact is inferior to no indexation at all,
  2. indexation to BWpred might provide a certain improvement compared to indexation to BWact and
  3. indexation according to height or no indexation at all (EVLW) are superior to indexation to BWact or BWpred.

These results are – at first glance – surprising, as several studies have suggested BWpred as the appropriate indexation factor [12]–[14].

Historically, different techniques have been established to quantify pulmonary edema termed as “EVLW” which was originally determined without indexation. Early studies frequently used animal models with post-mortem gravimetric determination of EVLW as the gold-standard [1], [ 2], [ 4], [ 10]. Regarding investigations in different species, indexation to BWact provided “basic” indexation allowing interspecies comparisons between different animal and human data.

However, in obese patients indexation to BWact might inappropriately diminish indexed EVLWIact. Based on a better prediction of mortality rather than on a better correlation to PaO2/FiO2, superiority of EVLWIpred to EVLWIact has been suggested: In the study by Phillips et al. [12] including 19 patients, EVLWIact was not related to mortality. By contrast, mortality of the seven out of 19 patients was univariately associated to EVLWIpred. However, EVLWIact obviously provided a better correlation to PaO2/FiO2 than EVLWIpred (coefficient of correlation: −0.525 for EVLWIpred and −0.773 for EVLWIact). With only 19 patients included, this study did not approach a multivariate mortality-analysis. Another study that included 44 patients (225 measurements) demonstrated better discrimination of ARDS- and non-ARDS-patients by EVLWIpred compared to EVLWIact [13]. A third study (44 patients; 44 measurements) showed the improved association of mortality by EVLWIpred compared to EVLWIact in a multivariate model [14]. However, similarly to the data of Phillips et al. this study did not demonstrate a better correlation of EVLWIpred to PaO2/FiO2 compared to EVLWIact (coefficients of correlation −0.57 vs. −0.55). This is in accordance with two more recent studies suggesting a comparable [16] or even stronger [15] association of EVLWIact compared to EVLWIpred with mortality.

Nevertheless, mortality is multifactorial, also depending on “do-not-resuscitate” statements and might be directly associated to some of the components of indexation (weight, BMI) [28]-[36]. E.g. further analysis of the data by Phillips et al. demonstrates that mean EVLW and BMI were increased to a similar degree in the non-survivors compared to survivors (45% and 31%, respectively). Therefore, mortality is not an obvious endpoint to compare the appropriateness of different indexations of EVLWI, particularly when applied in small mono-centric collectives.

In our study-groups un-indexed EVLW - next to EVLWIheight - provided the highest predictive capability regarding PaO2/FiO2 and OI. This indicates that particularly weight-based indexation might be a confounder rather than an improvement of the inter-individual comparison of EVLW(I) in an adult population. In the analysis of patients' means, PaO2/FiO2 was significantly associated to unindexed EVLW, EVLWIheight and EVLWIadj, whereas the correlation was not significant for all other indexations. These findings suggest that the association of EVLW(I) to PaO2/FiO2 might be obscured by inappropriate indexation. In general, the association of EVLW(I) was closer to OI compared to PaO2/FiO2. Including mean airway pressure in addition to PaO2/FiO2, OI also reflects the Positive End Expiratory Pressure (PEEP), peek and plateau pressure, ventilation mode (there is a usually higher mean airway pressure in controlled compared to assisted ventilation) as well as I:E ratio. Since the association of EVLW(I) to OI was not extensively investigated in the previous studies, the close association in our study might even strengthen the role of EVLW(I) as a parameter of pulmonary (patho)physiology.

With regard to indexation of other pulmonary parameters, the strong performance of height as an indexation for EVLWI is not surprising. As stated in recent consensus guidelines “lung volumes are related to body size, with standing height being the most important factor” [21].

Furthermore, a look at the “weight correction-formulas” demonstrates that BWpred and BWid (Table 1) do not contain any weight at all, but simply adjust height for gender and subtract a length-constant [20], [ 21], [ 28].

In addition to weight and height, the third major determinant of most indexation formulas is gender, which has impact on BWpred, BWid, BWadj and TLC. With regard to the above-mentioned formulas, indexation according to BWid increases EVLWI by 5.5% for women. EVLWIpred of women with a height between 150 and 190 cm is increased by 5–10% compared to men with the same height. However, this marked impact of gender on EVLW is not substantiated by our data: Multiple regression analysis regarding EVLW in our merged data including the variables age, height, weight, PaO2/FiO2 and gender demonstrated that gender was not independently associated to EVLW.

Finally, the question remains, whether in adults “no indexation at all” is the answer. Regarding our data in two adult groups with a high variability in body weight and BMI, but lower variability in height, this might be a reasonable option. However, it must be kept in mind that the variability in height was low in these groups: e.g. mean height in the Munich-evaluation-group was 170±10 cm (median 171 cm, range 150–190 cm).

On the other hand it is self-evident that indexation will improve the predictive capabilities in a group with a higher variability of parameters closely associated to EVLW such as height. There is elaborate data on the pulmonary function parameters in children and adolescents: Normal values in these groups with high variability in height and weight are mainly adjusted to height [20], [ 21].

Limitations of the study

Despite the inclusion of two different groups and the large sample size compared to the previous data, our study has several limitations. Our Munich-evaluation-group was a preselected group of non-operative mechanically ventilated patients with a prolonged ICU-stay. Although this drawback might be - at least in part - outweighed by a re-evaluation in a large group of non-selected anesthesiology patients, the data of both groups might not apply to patients without pulmonary impairment. On the other hand, the significance of modest correlations with r-values as low as -0.29 require cautious interpretation, since large numbers of patients promote significance of modest associations. Furthermore, these data are mainly derived from Caucasians. Despite a “considerable lack of data on lung-volumes in non-Caucasians” [20], [ 37] at least two studies give hints on differences regarding TLC between whites and blacks [38], Polynesians, Northern Indians and Pakistanis [39]–[41]. Finally, we cannot extrapolate our results to a pediatric population in which indexation to height may be much more appropriate than unindexed EVLW which was comparable to EVLWheight in our adult groups.

Conclusions

EVLW is a marker significantly associated to pulmonary function and mortality. Regarding the prediction of PaO2/FiO2 and OI, indexation of EVLWIact is inappropriate. EVLWIpred provides a slight improvement. The highest predictive capabilities in an adult population were found for EVLWIheight and un-indexed EVLW. Therefore, our data suggest that EVLW should be indexed to height (EVLWIheight) or remain unindexed in adults.

Acknowledgments

The authors would like to thank Analena Beitz and Benedikt Henschel for language revision.

Author Contributions

Conceived and designed the experiments: WH JH TS CC MM RS. Performed the experiments: WH JH AU MF BS CC MM. Analyzed the data: WH JH TS BS CC MM RS. Contributed to the writing of the manuscript: WH JH TS BS CC MM RS.

References

  1. 1. Neumann P (1999) Extravascular lung water and intrathoracic blood volume: double versus single indicator dilution technique. Intensive Care Med 25(2): 216–219.
  2. 2. Rossi P, Wanecek M, Rudehill A, Konrad D, Weitzberg E, et al. (2006) Comparison of a single indicator and gravimetric technique for estimation of extravascular lung water in endotoxemic pigs. Crit Care Med 34(5): 1437–1443.
  3. 3. Tagami T, Kushimoto S, Yamamoto Y, Atsumi T, Tosa R, et al. (2010) Validation of extravascular lung water measurement by single transpulmonary thermodilution: human autopsy study. Crit Care 14(5): R162.
  4. 4. Kuzkov VV, Kirov MY, Sovershaev MA, Kuklin VN, Suborov EV, et al. (2006) Extravascular lung water determined with single transpulmonary thermodilution correlates with the severity of sepsis-induced acute lung injury. Crit Care Med 34(6): 1647–1653.
  5. 5. Sakka SG, Klein M, Reinhart K, Meier-Hellmann A (2002) Prognostic value of extravascular lung water in critically ill patients. Chest 122(6): 2080–2086.
  6. 6. Eisenberg PR, Hansbrough JR, Anderson D, Schuster DP (1987) A prospective study of lung water measurements during patient management in an intensive care unit. Am Rev Respir Dis 136(3): 662–668.
  7. 7. Mitchell JP, Schuller D, Calandrino FS, Schuster DP (1992) Improved outcome based on fluid management in critically ill patients requiring pulmonary artery catheterization. Am Rev Respir Dis 145(5): 990–998.
  8. 8. Fernandez-Mondejar E, Castano-Perez J, Rivera-Fernandez R, Colmenero-Ruiz M, Manzano F, et al. (2003) Quantification of lung water by transpulmonary thermodilution in normal and edematous lung. J Crit Care 18(4): 253–258.
  9. 9. Groeneveld AB, Verheij J (2006) Extravascular lung water to blood volume ratios as measures of permeability in sepsis-induced ALI/ARDS. Intensive Care Med 32(9): 1315–1321.
  10. 10. Kirov MY, Kuzkov VV, Kuklin VN, Waerhaug K, Bjertnaes LJ (2004) Extravascular lung water assessed by transpulmonary single thermodilution and postmortem gravimetry in sheep. Crit Care 8(6): R451–458.
  11. 11. Martin GS, Eaton S, Mealer M, Moss M (2005) Extravascular lung water in patients with severe sepsis: a prospective cohort study. Crit Care 9(2): R74–82.
  12. 12. Phillips CR, Chesnutt MS, Smith SM (2008) Extravascular lung water in sepsis-associated acute respiratory distress syndrome: indexing with predicted body weight improves correlation with severity of illness and survival. Crit Care Med 36(1): 69–73.
  13. 13. Berkowitz DM, Danai PA, Eaton S, Moss M, Martin GS (2008) Accurate characterization of extravascular lung water in acute respiratory distress syndrome. Crit Care Med 36(6): 1803–1809.
  14. 14. Craig TR, Duffy MJ, Shyamsundar M, McDowell C, McLaughlin B, et al. (2010) Extravascular lung water indexed to predicted body weight is a novel predictor of intensive care unit mortality in patients with acute lung injury. Crit Care Med 38(1): 114–120.
  15. 15. Chew MS, Ihrman L, During J, Bergenzaun L, Ersson A, et al. (2012) Extravascular lung water index improves the diagnostic accuracy of lung injury in patients with shock. Crit Care 16(1): R1.
  16. 16. Mallat J, Pepy F, Lemyze M, Barrailler S, Gasan G, et al. (2012) Extravascular lung water indexed or not to predicted body weight is a predictor of mortality in septic shock patients. J Crit Care 27(4): 376–83.
  17. 17. Jozwiak M, Silva S, Persichini R, Anguel N, Osman D, et al. (2013) Extravascular lung water is an independent prognostic factor in patients with acute respiratory distress syndrome. Crit Care Med 41(2): 472–80.
  18. 18. Huber W, Mair S, Götz SQ, Tschirdewahn J, Siegel J, et al. (2013) Extravascular lung water and its association with weight, height, age, and gender: a study in intensive care unit patients. Intensive Care Med 39(1): 146–50.
  19. 19. Wolf S, Riess A, Landscheidt JF, Lumenta CB, Schürer L, et al. (2013) How to perform indexing of extravascular lung water: a validation study. Crit Care Med 41(4): 990–8.
  20. 20. Stocks J, Quanjer PH (1995) Reference values for residual volume, functional residual capacity and total lung capacity. ATS Workshop on Lung Volume Measurements. Official Statement of The European Respiratory Society. Eur Respir J 8(3): 492–506.
  21. 21. Wanger J, Clausen JL, Coates A, Pedersen OF, Brusasco V, et al. (2005) Standardisation of the measurement of lung volumes. Eur Respir J 26(3): 511–522.
  22. 22. Ogden CL, Carroll MD, Curtin LR, McDowell MA, Tabak CJ, et al. (2006) Prevalence of overweight and obesity in the United States, 1999–2004. JAMA: the journal of the American Medical Association 295(13): 1549–1555.
  23. 23. Beutler S, Schmidt U, Michard F (2004) Hemodynamic monitoring in obese patients: a big issue. Crit Care Med 32(9): 1981.
  24. 24. Camporota L, De Neef M, Beale R (2012) Extravascular lung water in acute respiratory distress syndrome: potential clinical value, assumptions and limitations. Crit Care 16(1): 114.
  25. 25. Cordemans C, De laet I, Van Regenmortel N, Schoonheydt K, Dits H, et al. (2012) Aiming for negative fluid balance in patients with acute lung injury and increased intra-abdominal pressure: A pilot study looking at the effects of PAL treatment. Ann Intensive Care 2(Supplem 1): S15.
  26. 26. Cordemans C, De laet I, Van Regenmortel N, Schoonheydt K, Dits H, et al. (2012) Fluid management in critically ill patients: The role of extravascular lung water, abdominal hypertension, capillary leak and fluid balance. AnnIntensive Care 2(Supplem 1): S1.
  27. 27. Saville DJ (1990) Multiple comparison procedures—the practical solution. Am Statistician 44: 174–180.
  28. 28. Pai MP, Paloucek FP (2000) The origin of the “ideal” body weight equations. Ann Pharmacother 34(9): 1066–1069.
  29. 29. Akinnusi ME, Pineda LA, El Solh AA (2008) Effect of obesity on intensive care morbidity and mortality: a meta-analysis. Crit Care Med 36(1): 151–158.
  30. 30. Sakr Y, Madl C, Filipescu D, Moreno R, Groeneveld J, et al. (2008) Obesity is associated with increased morbidity but not mortality in critically ill patients. Intensive Care Med 34(11): 1999–2009.
  31. 31. Naimark A, Cherniack RM (1960) Compliance of the respiratory system and its components in health and obesity. J Appl Physiol 15: 377–382.
  32. 32. Bedell GN, Wilson WR, Seebohm PM (1958) Pulmonary function in obese persons. J Clin Invest 37(7): 1049–1060.
  33. 33. Ray CS, Sue DY, Bray G, Hansen JE, Wasserman K (1983) Effects of obesity on respiratory function. Am Rev Respir Dis 128(3): 501–506.
  34. 34. Jones RL, Nzekwu MM (2006) The effects of body mass index on lung volumes. Chest 130(3): 827–833.
  35. 35. Brivet FG (2012) Pulmonary vascular permeability index should not be called extravascular lung water pulmonary blood volume ratio. Crit Care 16(2): 417.
  36. 36. O′Brien JM Jr, Phillips GS, Ali NA, Lucarelli M, Marsh CB, et al. (2006) Body mass index is independently associated with hospital mortality in mechanically ventilated adults with acute lung injury. Crit Care Med 34(3): 738–744.
  37. 37. Cook CD, Hamann JF (1961) Relation of lung volumes to height in healthy persons between the ages of 5 and 38 years. J Pediatr 59: 710–714.
  38. 38. O′Brien RJ, Drizd TA (1983) Roentgenographic determination of total lung capacity: normal values from a National Population Survey. Am Rev Respir Dis 1983 128(5): 949–952.
  39. 39. Miller GJ, Saunders MJ, Gilson RJ, Ashcroft MT (1977) Lung function of healthy boys and girls in Jamaica in relation to ethnic composition, test exercise performance, and habitual physical activity. Thorax 32(4): 486–496.
  40. 40. Pool JB, Greenough A (1989) Ethnic variation in respiratory function in young children. Respir Med 83(2): 123–125.
  41. 41. Rahman MA, Ullah MB, Begum A (1990) Lung function in teenage Bangladeshi boys and girls. Respir Med 84(1): 47–55.