Supplementary Appendix

Transcription

1 Supplementary Appendix This appendix has been provided by the authors to give readers additional information about their work. Supplement to: Gattinoni L, Caironi P, Cressoni M, et al. Lung recruitment in patients with the acute respiratory distress syndrome. N Engl J Med 2006;354:

2 SUPPLEMENTARY APPENDIX Lung recruitment in patients with Acute Respiratory Distress Syndrome Luciano Gattinoni*, M.D., F.R.C.P., Pietro Caironi*, M.D., Massimo Cressoni*, M.D., Davide Chiumello*, M.D., V. Marco Ranieri, M.D., Michael Quintel, M.D., Ph.D., Sebastiano Russo, M.D., Nicolò Patroniti, M.D., Rodrigo Cornejo, M.D., Guillermo Bugedo, M.D. * Istituto di Anestesia e Rianimazione, Fondazione IRCCS Ospedale Maggiore Policlinico, Mangiagalli, Regina Elena di Milano; Università degli Studi di Milano, Italy; Dipartimento di Anestesia, Azienda Ospedaliera S. Giovanni Battista-Molinette, Università degli Studi di Torino, Italy; Anaesthesiologie II, Operative Intensivmedizin, Universitatsklinikum Gottingen, Gottingen, Germany; Dipartimento di Medicina Perioperatoria e Terapia Intensiva, Azienda Ospedaliera S. Gerardo di Monza; Università degli Studi Milano-Bicocca, Italy; Departamentos de Anestesiologia y Medicina Intensiva, Facultad de Medicina, Pontificia Universidad Catolica de Chile, Santiago, Chile. 1) Additional methods (page 2 9) 2) Additional results (page 10 48)

3 Gattinoni et al. Supplementary Appendix, page 2 ADDITIONAL METHODS Approval of the study protocol and patient consent The study was approved by the Institutional Review Boards of each participating Institutions (Istituto di Anestesia e Rianimazione, Fondazione IRCCS Ospedale Maggiore Policlinico, Mangiagalli, Regina Elena di Milano, Università degli Studi di Milano, Italy; Dipartimento di Anestesia, Azienda Ospedaliera S. Giovanni Battista-Molinette, Università degli Studi di Torino, Italy; Anesthesiologie II, Operative Intensivmedizin, Universitatsklinikum Gottingen, Gottingen, Germany; Departmentos de Anestesiologia y Medicina Intensiva, Facultad de Medicina, Pontificia Universidad Catolica de Chile, Santiago, Chile). Patient consent was obtained according to the national regulations of each participating Institutions. As the patients were incompetent, patient consent in Italy was delayed according to the Italian regulations ( delayed consent [1]). The family was informed of the study (although not required by the law) and the study was performed. As soon as competent, each patient was fully informed on what had been done, and a written permission of using data collected was obtained. Enrollment To take into account the availability of the CT-scan facility and the sufficient time to stabilize the patients in order to perform the study in a safe condition, the experimental protocol did not include any time-limitation to perform the study after the diagnosis of ALI/ARDS. As an example, if the hemodynamic conditions of the patient were not stable, the study was postponed. Indeed, the time between the onset of ALI/ARDS and the study varied from 1 to 21 days (median 4 days, mean 5±5 days). Of note, the length of the time-period between the intubation and the day of the study did not appear to have any influence on the results observed in our study population (see pages 44 to 48 of the Supplementary Appendix). Comparison groups Data for comparison groups were obtained at five different Hospitals (Istituto di Anestesia e Rianimazione, Fondazione IRCCS Ospedale Maggiore Policlinico, Mangiagalli, Regina

4 Gattinoni et al. Supplementary Appendix, page 3 Elena di Milano, Università degli Studi di Milano, Italy; I Servizio di Anestesia e Rianimazione, Ospedale Luigi Sacco di Milano, Italy; Dipartimento di Medicina Perioperatoria e Terapia Intensiva, Azienda Ospedaliera S. Gerardo di Monza, Università degli Studi Milano- Bicocca, Italy; Dipartimento di Anestesia e Rianimazione, Ospedale Riuniti di Bergamo, Università degli Studi Milano-Bicocca, Italy; Dipartimento Emergenza Urgenza, Ospedale Civile di Legnano, Italy). From hospital databases, a total of 63 patients with lung injury other than ALI/ARDS, who underwent from 2001 and 2005 a whole lung CT-scan, and in which CT-scan data were available, were first selected. Among them, 62 patients were affected by pneumonia, and 1 patients by congestive heart failure. As not directly affected by a primary lung disease, the patient with congestive heart failure was excluded from the subsequent analysis. Among the remaining population, 28 patients were affected by bilateral pneumonia, and 34 patients by unilateral pneumonia, as diagnosed by the attending physician of the Emergency Unit. Within the group of patients with bilateral pneumonia, 15 patients had a PaO 2 /FIO 2 value greater than 300 mmhg, not presenting inclusion criteria for ALI/ARDS; in contrast, the remaining 13 patients did actually meet the criteria for ALI/ARDS diagnosis (PaO 2 /FIO 2 less than 300 mmhg), but were not classified as such by the attending physician of the Emergency Unit. Therefore, to avoid any possible confusion and potential overlapping between patients with ALI/ARDS, especially the less severe ones, and patients with bilateral pneumonia, only patients with unilateral pneumonia (n=34) were selected and included in the comparison group. To ascertain the diagnosis of unilateral pneumonia, the ratio between the amount of non-aerated lung tissue of the affected lung and the total amount of non-aerated lung tissue was calculated for each patients, i.e. both patients with bilateral and patients with unilateral pneumonia, and only patients with a ratio greater than 0.7 were at first included. The diagnosis of unilateral pneumonia was then re-confirmed in each patients by visual examination of the whole lung CT-scan. Data of patients with healthy lungs were obtained from the hospital database of the Istituto di Anestesia e Rianimazione, Fondazione IRCCS Ospedale Maggiore Policlinico, Mangiagalli, Regina Elena di Milano, Università degli Studi di Milano, Italy. Thirty-nine patients who underwent a whole lung CT-scan for general clinical assessment were retrieved and included in the comparison group.

5 Gattinoni et al. Supplementary Appendix, page 4 Calculations of physiological variables The following equations were used for the computation of physiological respiratory variables. 1) Dead-space fraction (percent of tidal-volume): PaCO2 PECO dead space = PaCO 2 2 where PaCO 2 is the arterial partial pressure of carbon dioxide, and PECO 2 is the mixed expired partial pressure of carbon dioxide. This variable was automatically computed by the CO 2 SMO monitor (Novametrix, Wallingford, CT), and was obtained in 48 patients, as only in these patients PECO 2 was measured. 2) Alveolar-dead space fraction (percent of tidal-volume): alveolar dead space = PaCO2 P PaCO ET 2 CO 2 where PaCO 2 is the arterial partial pressure of carbon dioxide, and PETCO 2 is the end-tidal partial pressure of carbon dioxide. This measurement was obtained in 55 patients, as only in these patients PETCO 2 was measured, and was automatically computed by the CO 2 SMO monitor (Novametrix, Wallingford, CT). 3) Right-to-left intrapulmonary shunt fraction (percent of cardiac output): shunt CcO fraction = CcO 2 2 CaO CvO 2 2 where CcO 2 is the capillary oxygen content, CaO 2 is the arterial oxygen content, and CvO 2 is the venous oxygen content.

6 Gattinoni et al. Supplementary Appendix, page 5 The capillary oxygen content (CcO 2 ) was computed as: [ P O * 0.003] *100% *1. 39 CcO A + 2 = 2 Hb where PAO 2 is the alveolar partial pressure of oxygen, and Hb is the blood concentration of hemoglobin. The arterial oxygen content (CaO 2 ) was computed as: [ PaO 0.003] Hb * *1. 39 CaO 2 = + SaO 2 * 2 where PaO 2 is the arterial partial pressure of oxygen, Hb is the blood concentration of hemoglobin, and SaO 2 is the arterial oxygen saturation. The venous oxygen content (CvO 2 ) was computed as: [ PvO 0.003] Hb * *1. 39 CvO 2 = + SvO 2 * 2 where PvO 2 is venous partial pressure of oxygen, Hb is the blood concentration of hemoglobin, and SvO 2 is the venous oxygen saturation. As only 22 patients had a pulmonary artery catheter, while 63 had a central venous catheter, in the 41 patients having only a central venous catheter blood gas values obtained from the central venous blood were used for the computation of the right-to-left intrapulmonary shunt as a surrogate of mixed venous blood values. 4) The respiratory-system compliance (ml/cm of H 2 O): respiratory system compliance = plateau V T pressure PEEP

7 Gattinoni et al. Supplementary Appendix, page 6 where V T is the tidal volume, plateau pressure is the inspiratory plateau pressure, and PEEP is the positive end-expiratory pressure. PEEP values were corrected for values of intrinsic PEEP, when detected during an end-expiratory pause [2].

8 Gattinoni et al. Supplementary Appendix, page 7 CT-scan image analysis Approach to the quantitative CT-scan analysis The CT scan measures the reduction of the radiation intensity upon passage through matter, which is called linear attenuation coefficient (µ) [3]. Through different mathematical algorithm, a given attenuation number µ is assigned to each voxel. The attenuation number depends on the energy of x-ray photons, as well as on density and the atomic number (Z) of material [4]. As a result, the attenuation is primarily determined by the density (mass/volume) of the tissue. In a given voxel the CT number is expressed as: CT 1000* = µ ( µ µ ) water water where CT is the mean CT number, µ is the linear attenuation coefficient of the given material, and µ water is the linear attenuation coefficient of water. Indeed, for practical purposes the CT number is a measure of density [5], i.e.: CT = 1000 Volume of gas ( Volume of gas + Volume of tissue) In the present paper, we refer to the tissue weight [6], i.e.: CT Tissue weight = 1 * Total volume 1000 We believe that referring to tissue instead of volume has several advantages, as what is subjected to stress and strain during mechanical ventilation is specifically the lung tissue, and not the lung volume, or the volume of lung regions with different degrees of aeration.

9 Gattinoni et al. Supplementary Appendix, page 8 CT-scan image processing The procedure to perform the quantitative analysis of each single CT-scan image was performed for all the ALI/ARDS patients, as well as for all the patients with either healthy lungs or unilateral pneumonia, at the Istituto di Anestesia e Rianimazione, Fondazione IRCCS Ospedale Maggiore Policlinico, Mangiagalli, Regina Elena di Milano. Each single CT-slice was considered as a region of interest. In each slice, the outlines of the lungs were manually countered, following the internal rib margin, the external mediastinal margin and the diaphragm profile. Pleural effusions, as well as large vessels of lung hilum, were excluded from the countered image. Before quantitative analysis, the accuracy of each single countered images was assessed by a Radiologist (B.F., from the Department of Radiology, Fondazione IRCCS Ospedale Maggiore Policlinico, Mangiagalli, Regina Elena di Milano), blinded as to the patient and airway pressure applied. Moreover, the repeatability of this procedure was tested in five patients: the difference in the quantitative analysis obtained was within 2 percent (data not shown) In each CT slice, we first determined the frequency distribution of the CT numbers of a given lung regions [5]. Then, we arbitrarily defined four lung compartments corresponding to different degrees of aeration, using the most adopted CT thresholds in the literature [6]: non-aerated lung tissue (with a density between +100 HU and 100 HU), poorly-aerated lung tissue (with a density between 101 HU and 500 HU), normally-aerated lung tissue (with a density between 501 HU and 900 HU), and hyper-inflated lung tissue (with a density between 901 HU and 1000 HU). In mathematical terms, for a given compartment: volume of compartment = ( number of voxels) volume of voxel compartment where volume of compartment is the volume of a specific compartment having a particular degree of aeration, (number of voxels) compartment is the number of voxels included in that specific compartment (having that specific degree of aeration), and volume of voxels is the volume of a voxel.

10 Gattinoni et al. Supplementary Appendix, page 9 We believe important to understand the difference between the computation of lung compartment volume and the computation of lung compartment tissue weight. It is evident that the tissue weight equals lung volume only in the non-aerated compartment, where the average CT is equal to zero. For the other compartments, the computation of tissue weight of a specific lung compartment is derived by using the average CT number of that specific compartment in the formula of the tissue weight computation. Definition of the potentially recruitable lung Among the possible different methods to compute lung recruitment, i.e. the recruitment to aeration of a portion of the lung parenchyma, we choose to define lung recruitment as the difference of the non-aerated lung tissue between 5 cm of H 2 O PEEP and 45 cm of H 2 O inspiratory airway pressure, and expressed as a proportion of the total lung weight. We chose this method as the damages of mechanical ventilation to the lungs are primarily induced by the formation of intra-parenchymal shear forces, that are more likely to develop when a collapsed tissue becomes de-collapsed. From an anatomical point of view, these modifications within the lung parenchyma are reflected by the transformation of non-aerated lung tissue to aerated lung tissue. We believe that other methods to calculate lung recruitment (such as the volume of gas entering the poorly-aerated lung compartment [7]), are very likely to be correlated with gas exchange modifications, but less correlated with the physical triggers of the ventilator-induced lung injury.

11 Gattinoni et al. Supplementary Appendix, page 10 ADDITIONAL RESULTS Overall population and higher versus lower amount of recruitable lung tissue Potentially recruitable lung and lung injury severity The proportion of non-aerated lung tissue measured at 5 cm of H 2 O PEEP, an index of the underlying severity of the lung injury, was widely variable and normally distributed (Figure 1 of the Supplementary Appendix), with a mean proportion equal to 37±16 percent of the total lung tissue (95 percent confidence interval, 33 to 41 percent; median 37 percent), corresponding to an absolute amount of 587±392 grams (95 percent confidence interval, 492 to 682 grams; median 476 grams). The amount of potentially recruitable lung (PRL) appears to be a function of the proportion of non-aerated lung tissue to the total lung weight measured at the baseline 5 cm of H 2 O PEEP (r 2 =0.46, P<0.001, Figure 2 of the Supplementary Appendix). Of note, the prevalence of patients that were dead at ICU discharge was greater among patients with either higher amount of PRL or higher proportion of non-aerated lung tissue at 5 cm of H 2 O PEEP, in comparison with that observed among patients either with lower amount or PRL or lower proportion of non-aerated lung tissue at baseline PEEP (solid circles, Figure 2 of the Supplementary Appendix). Similarly, the survival during the first 28 days of ICU admission observed in patients with a higher amount of PRL was significantly shorter than that observed in patients with a lower amount of PRL (P=0.02, Figure 3 of the Supplementary Appendix). PEEP trial and estimation of the potentially recruitable lung Several physiological respiratory variables recorded at 5 and 15 cm of H 2 O of PEEP were closely associated with the amount of PRL (Table 1 of the Supplementary Appendix). However, considering in particular their variation between 5 and 15 cm of H 2 O PEEP, the amount of PRL was significantly associated only with the variation of arterial oxygen saturation (P<0.001), respiratory-system compliance (P=0.003), right-to-left intrapulmonary shunt fraction (P=0.03),

12 Gattinoni et al. Supplementary Appendix, page 11 venous partial pressure of oxygen (P=0.004), and venous oxygen saturation (P=0.001), a higher amount of PRL being associated with greater variations. Table 2 of the Supplementary Appendix shows the sensitivity and specificity of tests set up by using combined physiological respiratory variables to predict patients with a higher amount of PRL, both as originally hypothesized and tested as post-hoc analysis. Patients with ALI/ARDS from pneumonia and patients with ALI/ARDS from sepsis As reported in the main text of the manuscript, patients with ALI/ARDS from pneumonia appeared to be more frequent among patients with a higher amount of PRL, while patients with ALI/ARDS from sepsis were more frequent among patients with a lower amount of PRL, in contrast with previous investigations [8,9]. To investigate for any possible explanation of these findings, we analyzed the baseline clinical characteristics and CT-scan variables of ALI/ARDS patients from either pneumonia or sepsis. Unexpectedly, patients with ALI/ARDS from pneumonia showed a greater baseline severity of lung injury, as detected by a lower PaO 2 /FIO 2 (P<0.001) and respiratory-system compliance (P=0.05), and a higher dead-space (P=0.02) and shunt fraction (P=0.001) than ALI/ARDS with sepsis (Table 3 of the Supplementary Appendix). Similarly, patients with ALI/ARDS from pneumonia showed a greater lung tissue weight (P<0.001) and proportion of non-aerated lung tissue (P=0.04), and a lower proportion of normally-aerated lung tissue (P<0.001) than patients with ALI/ARDS from sepsis (Table 4 of the Supplementary Appendix). Moreover, similar findings were observed when the analysis was limited to ALI/ARDS patients from pneumonia and ALI/ARDS patients from sepsis who underwent the CT-scan study protocol within 3 days from the diagnosis, thereby ruling out the possible effect of time on these results (data not shown). Therefore, it is very likely that pneumonia led to a greater severity of lung injury as compared to sepsis syndrome in ALI/ARDS patients, and that, as a consequence, patients with ALI/ARDS from pneumonia had a greater amount of PRL. Indeed, these findings appear to follow the general message of the main manuscript: the greater the severity of lung injury, the greater is the amount of PRL.

13 Gattinoni et al. Supplementary Appendix, page 12 Patients with unilateral pneumonia Thirty-four patients with unilateral pneumonia who underwent a whole lung CT-scan for diagnostic purposes were retrospectively selected and included in the study as comparison group. Among this group, 20 patients were spontaneously breathing and 14 patients were mechanically ventilated. As expected, patients mechanically ventilated showed a greater severity of their systemic illness, as indicated by the SAPS II score (P=0.01) and the FIO 2 clinically employed (P=0.04), and higher values of PaCO 2 (P=0.002), as compared to patients in spontaneous breathing (Table 5 of the Supplementary Appendix). In contrast, the two groups of patients appeared to be greatly comparable with regard to the lung CT-scan functional anatomy, as judged by the lung tissue weight, the proportion of non-aerated lung tissue and the proportion of normally-aerated lung tissue (Table 5 of the Supplementary Appendix).

14 Gattinoni et al. Supplementary Appendix, page 13 Figure 1 - Supplementary Appendix 11 ALI patients without ARDS ARDS patients frequency [no. of patients] / -5-5 / 00 / 5 5 / / / / / / / / / / / / / / / 80 non-aerated lung tissue [% total lung weight] Figure 1 Supplementary Appendix. The frequency distribution of non-aerated lung tissue recorded at 5 cm of H 2 O PEEP in the overall study population (n=68), expressed as a proportion of the total lung weight. Dashed columns represent patients classified as affected by acute lung injury without ARDS (PaO 2 /FIO 2 less than 300), while gray columns represent patients classified as affected by ARDS (PaO 2 /FIO 2 less than 200). The non-aerated lung tissue was defined as the lung tissue having a physical density at CT-scan image analysis between +100 HU and 100 HU, representing the portion of lung parenchyma which is consolidated and/or collapsed, i.e. the lung injury severity.

15 Gattinoni et al. Supplementary Appendix, page 14 Figure 2 - Supplementary Appendix the amount of potentially recruitable lung [% total lung weight] dead survived non-aerated lung tissue [% total lung weight] Figure 2 Supplementary Appendix. The amount of PRL as a function of the severity of the lung injury, as estimated by the proportion of non-aerated lung tissue at 5 cm of H 2 O PEEP, in the overall study population (n=68). Open circles represent patients surviving to discharge from the Intensive Care Units (ICUs, after 29±27 days), closed circles represent patients dying before ICU discharge. The dashed horizontal line represents the median value of the amount of PRL (9 percent of the total lung weight; 95 percent interval confidence, 8 to 14 percent). An exponential function (y = y 0 + a*exp(b*x)) was used to describe the relationship between the amount of PRL and the proportion of non-aerated lung tissue.

16 Gattinoni et al. Supplementary Appendix, page 15 Figure 3 - Supplementary Appendix survival [%] P= lower amount of potentially recruitable lung higher amount of potentially recruitable lung days after admission to Intensive Care Unit Figure 3 Supplementary Appendix. The survival observed during the first 28 days of ICU admission in patients with either a lower or a higher amount of PRL. The dashed line represents patients with a lower amount of PRL (n=34), while solid lines represents patients with a higher amount of PRL (n=34). By the 28 th day, 2 patients with a lower amount of PRL and 9 patients with a higher amount of PRL had died (6 vs. 26 percent, respectively, P=0.02).

17 Gattinoni et al. Supplementary Appendix, page 16 Table 1 of the Supplementary Appendix Physiological respiratory variables at different PEEP levels* Lower amount of PRL ( 9 percent) n = 34 Higher amount of PRL (> 9 percent) n = 34 P value Plateau pressure at 5 PEEP (cm of H 2 O) 18 ±3 21 ±4 <0.001 Plateau pressure at 15 PEEP (cm of H 2 O) 29 ±4 30 ± PaO 2 /FIO 2 at 5 PEEP (mm Hg) 194 ± ±60 <0.001 PaO 2 /FIO 2 at 15 PEEP (mm Hg) 245 ± ± Delta PaO 2 /FIO 2 (mm Hg) 51 ±53 67 ± PaO 2 at 5 PEEP (mm Hg) 87 ±22 71 ± PaO 2 at 15 PEEP (mm Hg) 111 ± ± Delta PaO 2 (mm Hg) 24 ±27 40 ± SaO 2 at 5 PEEP (%) 96 ±3 92 ±5 <0.001 SaO 2 at 15 PEEP (%) 97 ±2 97 ± Delta SaO 2 (%) 1 ±2 5 ±4 <0.001 PaCO 2 at 5 PEEP (mm Hg) 39 ±7 44 ± PaCO 2 at 15 PEEP (mm Hg) 39 ±7 45 ± Delta PaCO 2 (mm Hg) 0 ±3 1 ± Arterial ph at 5 PEEP 7.43 ± ± Arterial ph at 15 PEEP 7.42 ± ±

18 Gattinoni et al. Supplementary Appendix, page 17 Dead space at 5 PEEP (% of tidal-volume) 51 ±12 63 ± Dead space at 15 PEEP (% of tidal-volume) 53 ±12 63 ± Delta dead space (% of tidal volume) 2 ±5 0 ± Alveolar-dead space at 5 PEEP (% of tidal volume) 15 ±11 25 ± Alveolar-dead space at 15 PEEP (% of tidal volume) 15 ±12 23 ± Delta alveolar-dead space (% of tidal volume) -0 ±7-3 ± Respiratory-system compliance at 5 PEEP (ml/cm of H 2 O) 51 ±19 38 ± Respiratory-system compliance at 15 PEEP (ml/cm of H 2 O) 46 ±17 40 ± Delta respiratory-system compliance (ml/cm of H 2 O) -5 ±10 2 ± Shunt at 5 PEEP (% of cardiac output) 34 ±12 45 ± Shunt at 15 PEEP (% of cardiac output) 28 ±10 35 ± Delta shunt (% of cardiac output) -6 ±6-10 ± PvO 2 at 5 PEEP (mm Hg) 41 ±5 41 ± PvO 2 at 15 PEEP (mm Hg) 42 ± 7 45 ± Delta PvO 2 (mm Hg) 1 ±4 4 ± SvO 2 at 5 PEEP (%) 76 ±6 74 ±6 0.04

19 Gattinoni et al. Supplementary Appendix, page 18 SvO 2 at 15 PEEP (%) 76 ±8 78 ± Delta SvO 2 (%) 0 ±4 5 ± PvCO 2 at 5 PEEP (mm Hg) 44 ±7 48 ± PvCO 2 at 15 PEEP (mm Hg) 45 ±7 49 ± Delta PvCO 2 (mm Hg) 1 ±4 1 ± Venous ph at 5 PEEP 7.38 ± ± Venous ph at 15 PEEP 7.38 ± ± * Values are mean ±SD. PEEP denotes positive end-expiratory pressure values, PaO 2 the arterial partial pressure of oxygen, FIO 2 the inspired oxygen fraction, SaO 2 the arterial oxygen saturation, PaCO 2 the arterial partial pressure of carbon dioxide, PvO 2 the venous partial pressure of oxygen, SvO 2 the venous oxygen saturation, and PvCO 2 denotes the venous partial pressure of carbon dioxide. The delta of each variable was calculated as the difference between the values recorded at 15 cm of H 2 O PEEP minus the values recorded at 5 cm of H 2 O PEEP. P values were obtained by Student s t-test, and Wilcoxon s test analysis as appropriate. The dead space was calculated using a standard formula (see above). This measurement was available for 48 patients (23 patients with a lower, and 25 patients with a higher amount of PRL). The alveolar-dead space was calculated using a standard formula (see above). This measurement was available for 55 patients (27 patients with a lower, and 28 patients with a higher amount of PRL). Respiratory-system compliance was calculated as the ratio between the tidal volume and the difference between inspiratory plateau pressure and PEEP. The intrapulmonary right-to-left shunt was calculated using a standard formula (see above). This measurement was available for 60 patients (29 patients with a lower, and 31 patients with a higher amount of PRL).

20 Gattinoni et al. Supplementary Appendix, page 19 Table 2 of the Supplementary Appendix Sensibility, specificity, positive and negative predictive values of different tests to estimate patients with a higher amount of potentially recruitable lung* Sensitivity Specificity Positive Negative % % predictive value predictive value (no. of patients) (no. of patients) % % Delta PaO 2 /FIO 2 > 0 mm Hg, delta PaCO 2 < 0 mm Hg, delta respiratory-system compliance > 0 ml/cm of H 2 O Delta PaO 2 /FIO 2 > 0 mm Hg, delta alveolar-dead space < 0 % of tidal-volume, delta respiratory-system compliance > 0 ml/cm of H 2 O 71 (24/34) 59 (20/34) (24/28) 52 (14/27) PaO 2 /FIO 2 at 5 PEEP < 150 mmhg 74 (25/34) 76 (26/34) PaO 2 /FIO 2 at 5 PEEP < 150 mmhg, delta PaCO 2 < 0 mm Hg, delta respiratory-system compliance > 0 ml/cm of H 2 O PaO 2 /FIO 2 at 5 PEEP < 150 mm Hg, delta alveolar-dead space < 0 % of tidal-volume, delta respiratory-system compliance > 0 ml/cm of H 2 O 59 (20/34) 82 (28/34) (22/28) 81 (22/27) 81 79

21 Gattinoni et al. Supplementary Appendix, page 20 * PaO 2 denotes the arterial partial pressure of oxygen, FIO 2 the inspired oxygen fraction, PaCO 2 the arterial partial pressure of carbon dioxide, and PEEP denotes positive-end expiratory pressure levels. The delta of each variable was calculated as the difference between value recorded at 15 cm of H 2 O PEEP minus the values recorded at 5 cm of H 2 O PEEP. Respiratory-system compliance was calculated as the ratio between the tidal volume and the difference between inspiratory plateau pressure and PEEP. The alveolar-dead space was calculated using a standard formula (see above). This measurement was available for 55 patients (27 patients with a lower, and 28 patients with a higher amount of PRL). This test was considered positive when two or three of the variable variations indicated occurred when increasing PEEP from 5 to 15 cm of H 2 O.

22 Gattinoni et al. Supplementary Appendix, page 21 Table 3 of the Supplementary Appendix Baseline clinical characteristics and mortality rate of patients with ALI/ARDS from either pneumonia or sepsis at PEEP of 5 cm of H 2 O* Overall ALI/ARDS from pneumonia ALI/ARDS from sepsis P value n = 49 n = 25 n = 24 Age (yrs) 53 ±16 51 ±17 54 ± Female sex no. of patients (%) 31 (63) 17 (68) 10 (42) 0.06 Body mass index (kilogram/meters 2 ) 25 ±5 24 ±4 27 ± SAPS II 36 ±12 35 ±12 37 ± Tidal-volume (ml) 529 ± ± ± Minute ventilation (liters/min) 10.2 ± ± ± Respiratory rate (breaths/min) 20 ±7 21 ±7 18 ± PEEP (cm of H 2 O) 11 ±3 11 ±3 11 ± PaO 2 /FIO 2 (mm Hg) 163 ± ± ±75 <0.001 PaO 2 (mm Hg) 78 ±23 71 ±19 85 ± FIO 2 (%) 53 ±16 60 ±18 46 ± PaCO 2 (mm Hg) 42 ±9 44 ±10 40 ± ph 7.39 ± ± ± Dead space (% of tidal-volume) 58 ±15 64 ±13 52 ± Respiratory-system compliance (ml/cm of H 2 O) 41 ±15 37 ±15 45 ±

23 Gattinoni et al. Supplementary Appendix, page 22 Shunt (% of cardiac output) 40 ±17 48 ±18 33 ± Fluid balance before the study (ml/day)** 2477 ± ± ± Days of ventilation before the study 6 ±6 5 ±6 6 ± Intra-abdominal pressure (cm of H 2 O) 13 ±5 11 ±4 16 ± Mortality at ICU discharge (no. of patients, %) 15 (31) 9 (36) 6 (25) 0.40 * Values are mean ±SD. PEEP denotes positive end-expiratory pressure values, FIO 2 the inspired oxygen fraction, PaO 2 the arterial partial pressure of oxygen, and PaCO 2 denotes the arterial partial pressure of carbon dioxide. P values were obtained by Student s t- test, Wilcoxon s test, and Chi-square test analysis as appropriate. The Simplified Acute Physiology Score II (SAPS II) [10] was used to assess the severity of systemic illness at study entry. Possible scores range from 0 to 163, with higher scores indicating more severe illness. The dead space was calculated using a standard formula (see above). This measurement was available for 35 patients (19 patients with ALI/ARDS from pneumonia, and 16 patients with ALI/ARDS from sepsis). Respiratory-system compliance was calculated as the ratio between the tidal volume and the difference between inspiratory plateau pressure and PEEP. The intrapulmonary right-to-left shunt was calculated using a standard formula (see above). This measurement was available for 43 patients (20 patients with ALI/ARDS from pneumonia, and 23 patients with ALI/ARDS from sepsis). ** Fluid balance before the study averaged for each patient the daily fluid balance within the last five days before the study. Days of mechanical ventilation before the study were counted from the day of intubation (day 0) to the day of the study. The average time of discharge from ICU was 29 ±27 days, ranging from 2 to 163 days (median 22.5 days).

24 Gattinoni et al. Supplementary Appendix, page 23 Table 4 of the Supplementary Appendix Lung CT-scan variables of patients with ALI/ARDS from either pneumonia or sepsis at PEEP of 5 cm of H 2 O* Overall ALI/ARDS from pneumonia ALI/ARDS from sepsis P value n = 49 n = 25 n = 24 Potentially recruitable lung (grams) 230 ± ± ± Potentially recruitable lung (% of total lung weight) 13 ±13 19 ±14 8 ± Lung volume (ml) 2623 ± ± ± Lung tissue weight (grams) 1521 ± ± ±346 <0.001 Gas volume (ml) 1102 ± ± ± Non-aerated lung tissue (grams) 617 ± ± ± Non-aerated lung tissue (% of total lung weight) 38 ±17 44 ±17 32 ± Consolidated lung tissue (grams) 387 ± ± ± Consolidated lung tissue (% of total lung weight) 25 ±12 25 ±12 25 ± Poorly-aerated lung tissue (grams) 554 ± ± ± Poorly-aerated lung tissue (% of total lung weight) 35 ±12 38 ±13 32 ±

25 Gattinoni et al. Supplementary Appendix, page 24 Normally-aerated lung tissue (grams) 348 ± ± ± Normally-aerated lung tissue (% of total lung weight) 26 ±16 18 ±13 35 ±14 <0.001 Hyper-inflated lung tissue (grams) 2 ±11 4 ±15 1 ± Hyper-inflated lung tissue (% of total lung weight) 0 ±1 0 ±1 0 ± * Values are mean ±SD. Non-aerated lung tissue denotes the portion of lung parenchyma having density between +100 HU and 100 HU, consolidated lung tissue the portion of non-aerated lung tissue which remained non-aerated even at 45 cm of H 2 O airway pressure, poorly-aerated lung tissue the portion of lung parenchyma having a density between 101 HU and 500 HU, normallyaerated lung tissue the portion of lung parenchyma having a density between 501 HU and 900 HU, and hyper-inflated lung tissue denotes the portion of lung parenchyma having a density between 901 HU and 1000 HU. The potential for lung recruitment was defined as the portion of non-aerated lung tissue regaining aeration from 5 to 45 cm of H 2 O airway pressure. P values were obtained Student s t-test, and Wilcoxon s test as appropriate.

26 Gattinoni et al. Supplementary Appendix, page 25 Table 5 of the Supplementary Appendix Clinical characteristics and baseline lung CT-scan variables in patients with unilateral pneumonia during either spontaneous breathing or mechanical ventilation* Patients with unilateral pneumonia spontaneously breathing n = 20 Patients with unilateral pneumonia mechanically ventilated n = 14 P value Age 69 ±18 59 ± SAPS II 28 ±8 46 ± PaO 2 /FIO 2 (mm Hg) 235 ± ± FIO 2 (%) 35 ±15 54 ± SaO 2 (%) 93 ±4 89 ± PaCO 2 (mm Hg) 36 ±6 46 ± Arterial ph 7.44 ± ± Lung volume (ml) 3531 ± ± Lung tissue weight (grams) 1150 ± ± Gas volume (ml) 2382 ± ± Non-aerated lung tissue (grams) 291 ± ± Non-aerated lung tissue (% of total lung weight) 24 ±14 33 ± Poorly-aerated lung tissue (grams) 298 ± ± Poorly-aerated lung tissue (% of total lung weight) 25 ±9 22 ±7 0.62

27 Gattinoni et al. Supplementary Appendix, page 26 Normally-aerated lung tissue (grams) 527 ± ± Normally-aerated lung tissue (% of total lung weight) 48 ±12 45 ± Hyper-inflated lung tissue (grams) 34 ±55 6 ± Hyper-inflated lung tissue (% of total lung weight) 3 ±4 1 ± Aerated-lung tissue (grams) 859 ± ± Aerated-lung tissue (% of total lung weight) 76 ±14 67 ± Mortality at hospital discharge (no. of patients, %) 4 (20) 2 (14) 1.0 * Values are mean ±SD. PaO 2 denotes the arterial partial pressure of oxygen, FIO 2 the inspired oxygen fraction, SaO 2 the arterial oxygen saturation, PaCO 2 the arterial partial pressure of carbon dioxide, non-aerated lung tissue the portion of lung parenchyma having a density between +100 HU and 100 HU, poorly-aerated lung tissue the portion of lung parenchyma having a density between 101 HU and 500 HU, normally-aerated lung tissue the portion of lung parenchyma having a density between 501 HU and 900 HU, and hyper-inflated lung tissue the portion of lung parenchyma having a density between 901 HU and 1000 HU, and aerated lung tissue denotes the portion of lung parenchyma having a density between 101 HU and 1000 HU, i.e. the sum of poorly-aerated, normally-aerated, and hyper-inflated lung tissue. P values were obtained Student s t-test, Wilcoxon s test, and Fisher exact test analysis as appropriate. The Simplified Acute Physiology Score II (SAPS II) [10] was used to assess the severity of systemic illness at study entry. Possible scores range from 0 to 163, with higher scores indicating more severe illness. The average time of discharge from hospital was 25 ±21 days, ranging from 0 to 75 days (median 21 days).

28 Gattinoni et al. Supplementary Appendix, page 27 Analysis according to the quartile distribution of the amount of potentially recruitable lung To further characterize the distribution of the amount of PRL among our ALI/ARDS patient population, we divided the study population into for groups using the quartile values of the distribution of the amount of PRL: patients with a very low amount of PRL (less than 6 percent of the total lung weight), patients with a low amount of PRL (between 6 and 9 percent), patients with a high amount of PRL (between 9 and 19 percent), and patients with a very high amount of PRL (more than 19 percent of the total lung weight). Clinical characteristics The pre-study clinical characteristics of the four groups of patients were similar with regard to age, female sex prevalence, body mass index, severity of illness as assessed by the SAPS II score, daily fluid intake before the study, and days of mechanical ventilation before the study (Table 6 of the Supplementary Appendix). Tidal volume, PEEP and minute ventilation clinically employed were not different between the groups. In contrast, respiratory-system compliance and PaO 2 /FIO 2 were progressively lower in patients with higher amounts of PRL in comparison with patients with lower amount of PRL (P=0.009 and P=0.005, respectively, Table 6 of the Supplementary Appendix). Moreover, the mortality rate, both 28 days after ICU admission and at ICU discharge, progressively increased along with the amount of PRL (P=0.002 and P=0.006, respectively, Table 6 of the Supplementary Appendix). Bedside prediction of the amount of potentially recruitable lung To provide an estimation at the bedside of the amount of PRL, patients underwent a PEEP trial at two different PEEP levels, i.e. 5 and 15 cm of H 2 O. As already described in the main text of the manuscript, we initially hypothesized that patients with higher amounts of PRL would have shown at least two of the following respiratory variable responses when increasing PEEP from 5 to 15 cm of H 2 O: an increase in PaO 2 /FIO 2, a decrease in PaCO 2, and an increase in the respiratory-system compliance. Indeed, this test succeeded in distinguish patients with higher amounts of PRL from patients with lower amounts of PRL with an acceptable accuracy, especially for the extreme quartiles (patients with a very low vs. patients with a very high amount of PRL). In contrast, the power of this test in characterizing the patients within the two

29 Gattinoni et al. Supplementary Appendix, page 28 intermediate quartiles (patients with a low and patients with a high amount of PRL) was quite poor (Table 7 of the Supplementary Appendix). Considering the variation of alveolar-dead space in place of the variation of PaCO 2 increased the accuracy of detecting patients with a very high amount of PRL (100 percent of the patients detected, Table 7 of the Supplementary Appendix), but did not improve the characterization and the distinction between the two intermediate quartiles. The use of the PaO 2 /FIO 2 value at 5 cm of H 2 O PEEP in the place of its variation between 5 and 15 cm of H 2 O PEEP significantly improve the accuracy of the estimation of the amount of PRL. Patients with a PaO 2 /FIO 2 value at 5 cm of H 2 O PEEP less than 150 mm Hg were markedly more frequent within the two groups of patients with higher amounts of PRL as compared to the two groups of patients with lower amounts of PRL (Table 7 of the Supplementary Appendix). Among all the combinations of physiological respiratory variables tested, the best predictor of patients with higher amounts of PRL appeared to be the presence of at least two of the following parameters: a PaO 2 /FIO 2 value at 5 cm of H 2 O PEEP less than 150 mm Hg, a decrease in alveolar-dead space, and an increase in the respiratory-system compliance when increasing PEEP from 5 to 15 cm of H 2 O.

30 Gattinoni et al. Supplementary Appendix, page 29 Table 6 of the Supplementary Appendix Baseline clinical characteristics and mortality rate of the study population according to the quartile distribution of the amount of potentially recruitable lung* Very low Low High Very high amount of PRL amount of PRL amount of PRL amount of PRL ( 6 percent) (> 6 and 9 (> 9 and 19 (> 19 percent) P value percent) percent) n = 17 n = 17 n = 17 n = 17 Age (yrs) 57 ±17 55 ±16 60 ±18 47 ± Female sex no. of patients (%) 8 (47) 7 (41) 4 (24) 14 (83) 0.11 Body mass index (kilogram/meters 2 ) 26 ±5 25 ±4 26 ±5 23 ± SAPS II 38 ±11 35 ±13 34 ±8 39 ± Tidal-volume (ml/kg ideal body weight) 8.5 ± ± ± ± Minute ventilation (liters/min) 9.8 ± ± ± ± Respiratory rate (breaths/min) 19 ±7 16 ±5 17 ±6 21 ± PEEP (cm of H 2 O) 10.6 ± ± ± ± Plateau pressure (cm of H 2 O) 23 ±2 24 ±3 25 ±5 27 ± Respiratory-system compliance (ml/cm of H 2 O)** 50 ±15 48 ±16 46 ±21 33 ± PaO 2 /FIO 2 (mm Hg) 231 ± ± ± ± FIO 2 (%) 43 ±8 50 ±11 45 ±8 63 ±

31 Gattinoni et al. Supplementary Appendix, page 30 PaCO 2 (mm Hg) 38 ±7 38 ±10 42 ±11 49 ± Arterial ph 7.43 ± ± ± ± Causes of lung injury (no. of patients, %): Pneumonia 3 (18) 4 (24) 6 (35) 12 (71) Sepsis 10 (59) 7 (41) 5 (29) 2 (12) 0.03 Aspiration 2 (12) 1 (6) 1 (6) 0 (0) 0.90 Trauma 1 (6) 2 (12) 0 (0) 0 (0) 0.61 Others 1 (6) 3 (18) 5 (29) 3 (18) 0.40 Fluid balance before the study (ml/day) 1304 ± ± ± ± Days of ventilation before the study 5 ±5 6 ±7 6 ±7 6 ± ALI / ARDS (no. of patients) 9 / 8 5 / 12 5 / 12 0 / Mortality 28-days after ICU entry (no. of patients, %) 0 (0) 2 (12) 2 (12) 7 (41) Mortality at ICU discharge (no. of patients, %) 2 (12) 3 (18) 5 (29) 9 (53) * Values are mean ±SD. PEEP denotes positive end-expiratory pressure values, PaO 2 the arterial partial pressure of oxygen, FIO 2 the inspired oxygen fraction, PaCO 2 the arterial partial pressure of carbon dioxide, and ICU denotes Intensive Care Unit. Because of rounding, percentages may not total 100. P values were obtained by Student s t-test, Fisher exact test, Chi-square test and Mantel-Haenszel s test analysis as appropriate. P<0.01 vs. patients with a very low amount of PRL. P<0.01 vs. patients with a low amount of PRL.

32 Gattinoni et al. Supplementary Appendix, page 31 P<0.05 vs. patients with a high amount of PRL. The Simplified Acute Physiology Score II (SAPS II) [10] was used to assess the severity of systemic illness at study entry. Possible scores range from 0 to 163, with higher scores indicating more severe illness. ** Respiratory-system compliance was calculated as the ratio between the tidal volume and the difference between inspiratory plateau pressure and PEEP. Other causes of acute lung injury included anaphylactic shock, acute lung injury after surgery and following bone marrow transplantation. Fluid balance before the study averaged for each patient the daily fluid intake within the last five days before the study. Days of mechanical ventilation before the study were counted from the day of intubation (day 0) to the day of the study. The average time of discharge from ICU was 29±27 days, ranging from 2 to 163 days (median 22.5 days).

33 Gattinoni et al. Supplementary Appendix, page 32 Table 7 of the Supplementary Appendix Prediction of the amount of potentially recruitable lung according to its quartile distribution* Very low Low High Very high amount amount amount amount of PRL of PRL of PRL of PRL ( 6 percent) n = 17 (> 6 and 9 percent) n = 17 (> 9 and 19 percent) n = 17 (> 19 percent) n = 17 No. of patients with 2 or 3 of the following positive parameters (%): delta PaO 2 /FIO 2 > 0 mm Hg, delta PaCO 2 < 0 mm Hg, delta respiratory-system compliance > 0 ml/cm of H 2 O No. of patients with 2 or 3 of the following positive parameters (%): delta PaO 2 /FIO 2 > 0 mm Hg, delta alveolar-dead space < 0 % of tidal-volume, delta respiratory-system compliance > 0 ml/cm of H 2 O 5 (29) 9 (53) 10 (59) 14 (82) 3 (25) 10 (67) 9 (69) 15 (100) No. of patients with PaO 2 /FIO 2 at 5 PEEP < 150 mm Hg (%) 3 (18) 5 (29) 9 (53) 16 (94) No. of patients with 2 or 3 of the following positive parameters (%): PaO 2 /FIO 2 at 5 PEEP < 150 mm Hg, delta PaCO 2 < 0 mm Hg, delta respiratory-system compliance > 0 ml/ cm of H 2 O 2 (12) 4 (24) 6 (35) 14 (82)

34 Gattinoni et al. Supplementary Appendix, page 33 No. of patients with 2 or 3 of the following positive parameters (%): PaO 2 /FIO 2 at 5 PEEP < 150 mm Hg, delta alveolar-dead space < 0 % of tidal-volume, delta respiratory-system compliance > 0 ml/cm of H 2 O 1 (8) 4 (27) 8 (62) 14 (93) * PaO 2 denotes the arterial partial pressure of oxygen, FIO 2 the inspired oxygen fraction, PaCO 2 the arterial partial pressure of carbon dioxide, and PEEP denotes positive-end expiratory pressure levels. The delta of each variable was calculated as the difference between the values recorded at 15 cm of H 2 O PEEP minus the values recorded at 5 cm of H 2 O PEEP. Respiratory-system compliance was calculated as the ratio between the tidal volume and the difference between inspiratory plateau pressure and PEEP. The alveolar-dead space was calculated using a standard formula (see above). This measurement was available for 55 patients (27 patients with a lower, and 28 patients with a higher amount of PRL).

35 Gattinoni et al. Supplementary Appendix, page 34 Predictors of mortality To investigate any possible association between baseline respiratory and clinical variables and an increased risk of death in the overall study population, we first analyze the clinical, gas exchange, and respiratory mechanics parameters recorded at baseline 5 cm of H 2 O PEEP between patients who survived and patients who did not survive at discharge from ICU (29±27 days after ICU admission, Table 8 of the Supplementary Appendix), by using a univariate analysis (Student s t-test or Wilcoxon s test as appropriate). The variables significantly different between survivors and non-survivors were then grouped into two main categories: carbon dioxide-related variables (such as dead space, alveolar dead space, and CO 2 production), and other gas exchange-related variables (such as arterial ph, venous ph, and PvO 2 ). For each category, all the variables significantly different between survivors and non-survivors (P<0.05) were introduced into a stepwise, backward, multiple-logistic regression model. The SAPS II score and the heart rate at 5 cm of H 2 O PEEP were already considered for the subsequent analysis. The respiratory and clinical variables identified in the two categories as independently associated with an increased risk of death, as well as the SAPS II score and heart rate at 5 cm of H 2 O PEEP, were re-introduced into a final stepwise, backward, multiple-logistic regression model in order to identify among all the physiological respiratory and clinical variables the most independently associated with an increased risk of death. Among the carbon dioxide-related variables significantly different from survivors and nonsurvivors at ICU discharge, only dead space at 5 cm of H 2 O PEEP appeared to be independently associated with an increased risk of death (P=0.002). Among the other gas exchange-related variables, the only parameters independently associated with an increased risk of death was the arterial ph at 5 cm of H 2 O PEEP (P=0.01). The following physiological respiratory and clinical variables were therefore introduced into a stepwise, backward, multiple-logistic regression model: SAPS II score, dead space at 5 cm of H 2 O PEEP, arterial ph at 5 cm of H 2 O PEEP, and heart rate at 5 cm of H 2 O PEEP. Among them, the variables that appeared to be independently associated with an increased risk of death were SAPS II score (P=0.04), dead space at 5 cm of H 2 O PEEP (P=0.03), and heart rate at 5 cm of

36 Gattinoni et al. Supplementary Appendix, page 35 H 2 O PEEP (P=0.02, Table 9 of the Supplementary Appendix). The odds of death increased as SAPS II score increased, dead space at 5 cm of H 2 O PEEP increased, and as heart rate at 5 cm of H 2 O PEEP increased. The fit of the model was clearly good, as indicated by the Hosmer and Lemeshow goodness-of-fit test (P=0.96).

37 Gattinoni et al. Supplementary Appendix, page 36 Table 8 of the Supplementary Appendix Baseline respiratory and clinical variables associated with an increased risk of death* Survivors n = 49 Non-survivors n = 19 P value SAPS II 34 ±10 42 ± Dead space at 5 PEEP (% of tidal-volume) 53 ±12 67 ±12 <0.001 Alveolar-dead space at 5 PEEP (% of tidal-volume) 18 ±13 25 ± Heart rate at 5 PEEP (beats/min) 86 ± ±17 <0.001 Arterial ph at 5 PEEP 7.41 ± ± Venous ph at 5 PEEP 7.38 ± ± PvO 2 at 5 PEEP (mm Hg) 40 ±5 44 ± * Values are mean ±SD. PEEP denotes positive-end expiratory pressure values, and PvO 2 the venous partial pressure of oxygen. P values were obtained by Student s t-test or Wilcoxon s test analysis as appropriate. The dead space was calculated using a standard formula (see above). This measurement was available for 48 patients (23 patients with a lower, and 25 patients with a higher amount of PRL). The alveolar-dead space was calculated using standard formula (see above). This measurement was available for 55 patients (27 patients with a lower, and 28 patients with a higher amount of PRL).

38 Gattinoni et al. Supplementary Appendix, page 37 Table 9 of the Supplementary Appendix Odds ratios for the respiratory and clinical variables independently associated with an increased risk of death* Odds ratio (95% CI) P value SAPS II (per 1-point increase) 1.12 ( ) 0.04 Dead space at 5 PEEP (per increase of 5 percent) 1.51 ( ) 0.03 Heart rate at 5 PEEP (per increase of 5 beats/min) 1.40 ( ) 0.02 * Results were calculated by using a stepwise, backward, multiple-logistic regression. The odds of death increased as SAPS II score increased, the dead space at 5 cm of H 2 O PEEP increased, and the heart rate at 5 cm of H 2 O PEEP increased (Hosmer and Lemeshow goodness-of-fit test, P=0.9577, c=0.921). CI denotes confidence interval, SAPS II the Simplified Acute Physiology Score II and PEEP denotes positive end-expiratory pressure values. The dead space was calculated using a standard formula (see above). This measurement was available for 48 patients (23 patients with a lower, and 25 patients with a higher amount of PRL).