The mortality from acute respiratory distress syndrome after pulmonary resection is reducing: a 10-year single institutional experience

European Journal of Cardio-thoracic Surgery 34 (2008) 898 902 www.elsevier.com/locate/ejcts The mortality from acute respiratory distress syndrome after pulmonary resection is reducing: a 10-year single institutional experience Sarah S.K. Tang a, Karen Redmond a, Mark Griffiths b, George Ladas a, Peter Goldstraw a, Michael Dusmet a, * a Department of Thoracic Surgery, Royal Brompton Hospital, Sydney Street, London, SW3 6NP, UK b Adult Intensive Care Unit, Royal Brompton Hospital, Sydney Street, London, SW3 6NP, UK Received 29 February 2008; received in revised form 26 May 2008; accepted 9 June 2008 Abstract Objective: Acute respiratory distress syndrome (ARDS) is a major cause of death following lung resection. At this institution we reported an incidence of 3.2% and a mortality of 72.2% in a review of patients who underwent pulmonary resection from 1991 to 1997 [Kutlu C, Williams E, Evans E, Pastorino U, Goldstraw P. Acute lung injury and acute respiratory distress syndrome after pulmonary resection. Ann Thorac Surg 2000;69:376 80]. The current study compares our recent experience with this historical data to assess if improved recognition of ARDS and treatment strategies has had an impact on the incidence and mortality. Methods: We identified and studied all patients who developed ARDS following a lung resection of any magnitude between 2000 and 2005 using the 1994 consensus definition: characteristic chest X-ray or CT, PaO 2 / FiO 2 <200 mmhg, pulmonary capillary wedge pressure <18 mmhg and clinical acute onset. Overall incidence and mortality were recorded. Univariate analyses (t-test or x 2, as appropriate) were carried out to identify correlations between pre-, peri- and postoperative variables and outcomes. Results: We performed 1376 lung resections during the study period. Of these 705 (51.2%) were for lung cancer and 671 (48.8%) for other diseases. Twenty-two patients fulfilled the criteria for ARDS with 10 deaths in this group. The incidence and mortality from ARDS had fallen significantly over the two study periods (incidence from 3.2% to 1.6%, p = 0.01; mortality from 72% to 45%, p = 0.05). Although no significant correlations with incidence and mortality were identified, we found a number of significant trends. In keeping with the ARDS network study recommendations, postoperative tidal volumes were maintained at a lower level when a higher number of pulmonary segments were excised ( p = 0.001). Furthermore, consistent with findings in previous studies, the highest incidence and death from ARDS were in pneumonectomy patients (incidence 11.4%; mortality 50%). Although the incidence and mortality from ARDS following pneumonectomy were not significantly different between the two study periods ( p = 0.08, p = 0.35), we found that fewer pneumonectomies were performed in the later period (pneumonectomy rate of 6.4% vs 17.4%). Conclusions: The incidence and mortality of ARDS have decreased in our institution. We postulate that this is due to more aggressive strategies to avoid pneumonectomy, greater attention to protective ventilation strategies during surgery and to the improved ICU management of ARDS. # 2008 European Association for Cardio-Thoracic Surgery. Published by Elsevier B.V. All rights reserved. Keywords: ARDS; Resection; Incidence; Mortality; Protective ventilation 1. Introduction Acute respiratory distress syndrome (ARDS) remains a major cause of death following major lung resection, with a reported mortality rate ranging from 50% to 100% [1 4]. Prior to 1994, the lack of a concise definition for ARDS led to under reporting in patients suffering pulmonary complications after resection. This has been addressed by the publication of the 1994 consensus definition, outlining criteria for ARDS (Table 1). In 2000 this institution reported a 3.2% incidence of ARDS after lung resection with a mortality of 72.2%, with an Presented at the 21st Annual Meeting of the European Association for Cardio-thoracic Surgery, Geneva, Switzerland, September 16 19, 2007. * Corresponding author. Tel.: +44 207 3528121x8228; fax: +44 207 3528560. E-mail address: M.Dusmet@rbht.nhs.uk (M. Dusmet). increased incidence in male patients over the age of 60 years undergoing resection for lung cancer [1]. On the basis of this previous experience we have strived to decrease the pneumonectomy rate and to apply principles of protective ventilation strategies. Our understanding of the management of patients with ARDS has improved. This study was implemented to see if these factors have had an impact on either the incidence or the mortality of ARDS in our institution. 2. Materials and methods This is a retrospective review of a consecutive adult patient population (18 years and older) who underwent any lung resection for any pathology between January 2000 and May 2005. Patients who developed ARDS were identified from 1010-7940/$ see front matter # 2008 European Association for Cardio-Thoracic Surgery. Published by Elsevier B.V. All rights reserved. doi:10.1016/j.ejcts.2008.06.020

S.S.K. Tang et al. / European Journal of Cardio-thoracic Surgery 34 (2008) 898 902 899 Table 1 The 1994 consensus defining criteria for ARDS Table 2 Basic descriptors of the preoperative variables of the cohort with ARDS (n = 22) Clinical course Chest X-ray/CT PaO 2 /FiO 2 Exclusion of LV failure Acute onset Bilateral pulmonary infiltrates <200 mmhg PCWP <18 mmhg and/or normal echocardiogram Preoperative Mean SD Range PaO 2 (mmhg) 10.24 20.3 6 14 Sats (%) 95.86 1.94 92 99 FEV 1 (%) 70.54 24.32 18 115 FVC (%) 89.87 20.71 50 137 TLCO (%) 66.04 24.95 37 141 a thoracic adult intensive care database. We also reviewed the medical records of all patients who had an ICU stay in excess of 6 days and all those who died following pulmonary resection. The 1994 consensus definition for ARDS (Table 1) was applied. The incidence and mortality of ARDS following each type of resection were also recorded. The overall incidence and mortality were compared to figures from the previous review of patients who underwent pulmonary resection from 1991 to 1997 at our institution with a p value of <0.05 being considered significant (see Table 3). Multiple variables were analysed. The preoperative variables were: age, sex, cardiac comorbidity, pathology, number of segments excised, BMI, smoking history, administration of neoadjuvant chemotherapy, preoperative PaO 2 / FEV 1 /TLCO, ASA grade and predicted postoperative FEV 1 / TLCO. The intra-operative variables were length of operation, if systematic nodal dissection (SND) was performed and intra-operative fluid balance, including blood transfusion. Postoperative variables were: ventilation strategies (prone, positive end expiratory pressure (PEEP), tidal volume (TV), mode of ventilation, permissive hypercapnia), PaO 2 /FiO 2, arterial blood gases and whether steroids were administered. Haematological indices included highest white cell count and C-reactive protein. Sputum (or bronchoalveolar lavage) microbiology results were studied. ICU outcomes we studied included length of ICU stay, insertion of tracheostomy and mortality. Univariate analyses (t-test or x 2, as appropriate) were carried out to identify correlations between the named variables and objective outcomes. A two-tailed p value of <0.05 was considered statistically significant. Operative management: Rigid bronchoscopy followed by double lumen endotracheal intubation is routinely performed in this unit. Collapse of the ipsilateral lung allows dissection of the hilar elements, SND or precision excision of metastases as well as VATS surgery. The lungs are re-inflated prior to drainage and closure. Patients are then transferred to either a dedicated recovery unit or to the adult intensive care unit. Any patient who developed ARDS after lung resection and died, or that died within 30 days or in hospital following surgery was counted as a death. 3. Results Between January 2000 and May 2005, 1376 lung resections as defined above were performed. Of these 705 (51.2%) were for lung cancer and 671 (48.8%) for other diseases. Altogether there were 45 extended resections (for example lobectomy with en-bloc chest wall resection), 88 pneumonectomies, 601 lobectomies and 642 minor resections (including metastatectomies). Patients who underwent lung volume reduction surgery (LVRS) were also included. Twenty-two patients fulfilled the ARDS-defining criteria (Table 1) with an overall incidence of 1.6%. The following data characterise the study group of the 22 patients who developed ARDS. There were 16 men and 6 women with a mean age of 61 (range 29 82) years, with 11 patients over the age of 60 years. The mean BMI was 25 (range 17 34). Fourteen patients underwent surgery for lung cancer and eight for other diseases. Seventeen were either current or ex-smokers and five had never smoked. Their lung function test results are presented in Table 2. This demonstrates a wide range for all parameters as expected given the wide range of pathologies and procedures. The predicted postoperative value was calculated to take into account the amount of functional lung tissue resected [5]. Five patients had known co-existing cardiac disease (four had coronary artery disease of whom one had previous coronary bypass surgery and one patient had new atrial fibrillation diagnosed at preoperative assessment). Seven patients had received neoadjuvant chemotherapy. The median duration of anaesthesia in the ARDS group was 4.9 h (range 3 8 h). Ten out of 22 underwent right-sided resections, 10/22 left-sided resections and 2/22 had bilateral pulmonary resections. Out of the 22 patients, 15 underwent systemic nodal dissection. The mean operative fluid balance was 2.1 l (range 0.9 6.4 l), with three patients requiring intraoperative blood transfusions. None of these perioperative variables were shown to significantly correlate with outcome. Postoperatively, 19 of the 22 patients were treated with invasive ventilation with 16 patients requiring tracheostomy. Three patients were managed with non-invasive ventilation (one with CPAP and two with BIPAP). One patient was nursed prone. Steroids were administered to 13 patients. The mean FiO 2 requirement to maintain a mean postoperative PaO 2 of 9.4 kpa (1.5 kpa) was 60.3 kpa (18.7 kpa). The mean PEEP was 7 mmhg (2.7 mmhg) and the mean tidal volume was 526 ml (110 ml). A strategy of permissive hypercapnia was employed in 16 of 22 patients. Of note, the postoperative tidal volumes were maintained at a lower level when a higher number of pulmonary segments were excised ( p = 0.001; Fig. 1), in keeping with the ARDS network study recommendations [6]. Fifteen patients had echocardiography in the ICU. This showed poor right ventricular function in six and moderately poor left ventricular function in three. Severe right ventricular dysfunction in association with fulminant ARDS in three patients necessitated support with arteriovenous extracorporeal membranous oxygenation. Out of these patients, two survived to hospital discharge. The culture and sensitivities of sputum and/or bronchoalveolar lavage samples were all negative at the time of diagnosis of ARDS. The differential diagnosis of bacterial pneumonia was therefore excluded. Subsequent to the diagnosis of ARDS, cultures did show Gram-negative bacteria in six, a fungus in three and MRSA in four patients. The

900 S.S.K. Tang et al. / European Journal of Cardio-thoracic Surgery 34 (2008) 898 902 Fig. 1. Protective ventilatory management: this scatter plot shows that as the number of segments increases, the tidal volume decreases. remaining nine patients did not have reports of a positive culture. The hospital prophylaxis guidelines recommend the administration of 750 mg of Cefuroxime q8h intravenously for 24 h, the initial dose given 30 min prior to induction of general anaesthesia. The mean and median ICU length of stay were 31 (range 3 90) and 29 days, respectively. Patients requiring tracheostomy were more likely to have a longer postoperative ICU stay (39 days vs 6 days, p = 0.03). No preoperative, perioperative or postoperative variables were shown to significantly correlate with mortality in the ARDS patients. ARDS developed in patients undergoing all types of surgery for both benign and malignant disease (Table 4). Ten patients with ARDS died giving an overall mortality of 0.7% and a mortality of 45% within the ARDS group. Of those who died, five patients underwent pneumonectomy, three lobectomy and two LVRS (Table 4). There was no statistically significant relationship between extent of resection and mortality but the highest incidence (11.4%) and mortality (50%) were seen in pneumonectomy patients. When compared to our previous study [1], within the pneumonectomy group the incidence and mortality from ARDS was not significantly different between the two study periods ( p = 0.08, p = 0.5) but we found that fewer pneumonectomies were performed in the later study (6.4% vs 17.4% of all lung resections). 4. Discussion In 1967, Ashbaugh et al. first defined ARDS in a series of 12 patients with tachypnea, hypoxemia resistant to oxygen, diffuse alveolar infiltrates with reduced pulmonary compliance [7]. Over the following two decades, a lack of consensus in definition resulted in the use of different criteria to enrol patients in ARDS studies. A unifying definition of ARDS was established by the 1994 American-European Consensus Conference on ARDS [8]. It too has limitations that include poorly specified X-ray criteria that are subject to individual interpretation [9]. Although data applying the 1994 consensus definition for ARDS post-lung resection are limited, we used these criteria for our previous study in which we reported an incidence of 3.2% with an associated mortality of 72.2% [1]. In other studies the incidence of post-lung resection ARDS ranges from 1.4% to 3.6% with a mortality of 40 88% [2 4], with consistently higher incidence rates in patients undergoing more extensive resections (pneumonectomy 3.8 15%, lobectomy 2 5.6%, lesser resections 0.88 3.2%) [1 4]. The current study shows that both the overall incidence of post-lung resection ARDS and its mortality have fallen in our institution (incidence from 3.2% to 1.6%, p = 0.01; overall mortality from 2.3% to 0.7% and mortality in the ARDS group from 72% to 45%, p = 0.05, Table 3). Improved understanding of the pathophysiology of ARDS may account for the reduced incidence in our current review. It is generally accepted that activation of inflammatory cascades plays a pivotal role in ARDS [10 12]. However the mean highest white cell count of 27.4 8.7, potentially a surrogate marker of inflammation, did not correlate with outcome measures. Single lung ventilation, high-inspired oxygen concentrations, ischaemia-reperfusion resulting from collapse with re-inflation of the operative lung and mediastinal shift with hyperinflation of the remaining lung [13,14] are thought to potentially initiate this inflammatory cascade. Surgical manipulation may cause parenchymal injury similar to contusion, again activating the inflammatory cascade. Interestingly, in our study the length of operation (median 4.5, range 3 8 h), did not correlate with outcome. Although the ventilatory parameters on one lung ventilation are not recorded, the ventilatory strategies at this institution have become more protective over the past 5 years and a survey of in house members of the European Association of Cardiothoracic Anaesthesiologists is currently underway. Protective ventilatory strategies aim to minimise the risk of ventilator induced trauma to the lungs. In severely emphysematous lungs there is relatively little elastic tissue and they could therefore be more susceptible to volutrauma. This may explain the high incidence and mortality in the group of patients undergoing lung volume reduction surgery (Table 4). As discussed above, there is a body of evidence which suggests that pneumonectomy is associated with a Table 3 Comparison of incidence and deaths in our institution from ARDS in our previous study as reported in 2000 (1) and in the current study (2000 2005) 1991 1997 2000 2005 No. No. (incidence %) Deaths (mortality %) No. No. (incidence %) Deaths (mortality %) Extended resections 54 7 (13) 4 (57) 45 0 0 Pneumonectomy 198 10 (5) 8 (80) 88 10 (11.4) 5 (50) Lobectomy 612 17 (2.8) 12 (71) 601 6 (1) 3 (50) Minor resections 275 2 (0.7) 2 (100) 642 6 (0.9) 2 (33) Total 1139 36 (3.2) 26 (72) 1376 22 (1.6) 10 (45) p = 0.01 p = 0.05

S.S.K. Tang et al. / European Journal of Cardio-thoracic Surgery 34 (2008) 898 902 901 Table 4 Incidence and death from ARDS according to pathology and resection type Pneumonectomy (deaths/total) Lobectomy (deaths/total) Segmentectomy or lesser resection (deaths/total) LVRS (deaths/total) Overall (deaths/total) Lung cancer 3/8 3/6 0/0 0/0 6/14 Extrapulmonary malignancy 1/1 0/0 0/2 0/0 1/3 Benign lung disease 1/1 0/0 0/1 0/0 1/2 Emphysema 0/0 0/0 0/0 2/3 2/3 Overall (deaths/total) 5/10 3/6 0/3 2/3 10/22 higher incidence of post-lung resection ARDS. In the present series there were only 88 pneumonectomies (6.4%) as opposed to a pneumonectomy rate of 17.4% in our previous series. The employment of aggressive strategies to avoid pneumonectomy (e.g. improved preoperative staging including the use of PET and video mediastinoscopy, neoadjuvant chemotherapy to downstage patients and the employment of sleeve resections) could be one factor in the reduced incidence of ARDS at our institution (Table 3). In the recently published 2008 First National Thoracic Surgery Activity and Outcomes Report [15], pneumonectomy rates for primary lung cancer are between 3 and 34% in the UK (with an average of around 15%). This report also shows that the Royal Brompton Hospital has the highest rate of sleeve resections for primary lung cancer. Our unit has a sleeve resection rate (including angioplastic sleeve resections) of 10% whilst the majority of other UK units have rates between 0 and 4.5% with two other outliers at 6 and 8%. The role of fluid balance in the pathogenesis of ARDS is controversial [2,3,16,17]. SND could theoretically play a role in the development of ARDS with alteration in the normal lymphatic drainage promoting interstitial and alveolar oedema. SND is considered more extensive through a right thoracotomy, perhaps contributing to the higher incidence of ARDS seen after right, as opposed to left pneumonectomy [17 19]. However right pneumonectomy also removes a relatively larger amount of lung parenchyma and neither side of procedure or SND were statistically shown to affect outcome in our ARDS patients. Furthermore we do not perform SND when operating on benign disease, when performing VATS resections nor for routine metastasectomy. During mechanical ventilation PEEP helps by inflating and recruiting collapsed regions of the remaining lung tissue. However some authors feel that preferential ventilation of the relatively normal alveoli may result in over distension and lung injury [9,13,20]. Recently, the ARDS network study [5] found a significant reduction in mortality in ICU patients treated by lower tidal volumes (6 ml/kg vs 12 ml/kg). Similarly, another study found lower levels of inflammatory mediators in the bronchoalveolar lavage fluid in humans treated with low volume ventilation [21]. Treatment endpoints now focus on reducing peak inspiratory pressures (PIP) with higher PEEP whilst maintaining adequate oxygenation but specifically avoiding the maintenance of normal PaCO 2 (so-called permissive hypercapnia) [22]. Other ventilation strategies have been investigated in the past. Inverse ratio ventilation (IRV replaces the normal inspiratory to expiratory ratio of 1:3 with values of 2:1, 3:1 or 4:1) improves oxygenation by keeping alveoli open, recruiting further alveolar units and decreasing dead space does not also seem to have a survival benefit [23]. By recruiting previously collapsed regions of lung, nursing prone patients can markedly improve oxygenation but have not been shown to improve survival [24]. These strategies have been progressively introduced in our ICU as they have come to the fore. The use of steroids remains controversial with reports of increased mortality compared with placebo and recognised complications such as sepsis, pneumonia, wound infection, gastric ulceration and diabetes. Administration of steroids in late ARDS may limit fibrotic scarring however [25]. The results of the NIH ARDS Network Late Steroid Rescue Study are awaited. The use of steroids in our patient population reflects the trends of this debate. In this series veno-arterial ECMO was used in three patients with fulminant ARDS and associated right ventricular dysfunction resulted in two patients surviving to hospital discharge. 5. Conclusion In conclusion, the overall incidence (1.6%) and mortality (45%) of post-resection ARDS has fallen to roughly one half of their previous values in our institution over the past 15 years (Table 3). This probably reflects changes in treatment strategies, such as employing more aggressive techniques to avoid pneumonectomy and our greater appreciation of intraoperative protective ventilation strategies. Once ARDS does occur we have applied the lessons from our better understanding of the management of these patients in the ICU and this has had a great impact on improving our survival rates. Acknowledgement We would like to thank David Heavy, lecturer in research methods and biostatistics at Trinity College, Dublin, for the statistical analyses performed in this study. References [1] Kutlu C, Williams E, Evans E, Pastorino U, Goldstraw P. Acute lung injury and acute respiratory distress syndrome after pulmonary resection. Ann Thorac Surg 2000;69:376 80. [2] Dulu A, Pastores SM, Park B, Riedel E, Rusch V, Halpern NA. Prevalence and mortality of acute lung injury and ARDS after lung resection. Chest 2006;130(1):73 8. [3] Hayes JP, Williams EA, Goldstraw P, Evans TW. Lung injury in patients following thoracotomy. Thorax 1995;50:990 1. [4] Ruffini E, Parola A, Papalia E, Filosso PL, Mancuso M, Oliaro A, Actis-Dato G, Maggi G. Frequency and mortality of acute lung injury and acute

902 S.S.K. Tang et al. / European Journal of Cardio-thoracic Surgery 34 (2008) 898 902 respiratory distress syndrome after pulmonary resection for bronchogenic carcinoma. Eur J Cardiothorac Surg 2001;20:30 7. [5] Beckles MA, Spiro SG, Colice GL, Rudd RM. The physiologic evaluation of patients with lung cancer being considered for resectional surgery. Chest 2003;123:105S 14S. [6] Brower RG, Matthay M, Morris A, Schonfield D, Taylor Thompson B, Wheeler A, The Acute Respiratory Distress Syndrome Network. Ventilation with lower tidal volumes as compared with traditional tidal volumes for acute lung injury and the acute respiratory distress syndrome. N Eng J Med 2000;34(18):1301 8. [7] Ashbaugh DG, Bigelow BD, Petty TL, Levine BE. Acute respiratory distress in adults. Lancet 1967;ii:319 23. [8] Bernard GR, Artigas A, Brigham KL, Carlet J, Falke K, Hudson L, Lamy M, Legall JR, Morris A, Spragg R, The American-European Consensus Conference on ARDS. Definitions, mechanisms, relevant outcomes and clinical trial coordination. Am J Respir Crit Care Med 1994;149:818 24. [9] Rubenfeld GD, Caldwell E, Granton J, Hudson LD, Matthay MA. Interobserver variability in applying a radiographic definition for ARDS. Chest 1999;116:1347 53. [10] Pittet JF, Mackersie RC, Martin TR, Matthay MM. Biological markers of acute lung injury: prognostic and pathogenic significance. Am J Resp Crit Care Med 1997;155:1187 205. [11] Quinlan GJ, Evans TW, Gutteridge JMC. Linoleic acid and protine thiol changes suggestive of oxidative damage in the plasma of patients with the adult respiratory distress syndrome. Free Radic Res Commun 1994;20: 299 306. [12] Williams EA, Quinlan GJ, Goldstraw PG, Gothard JW, Evans TW. Postoperative lung injury and oxidative damage in patients undergoing pulmonary resection. Eur Respir J 1998;11:1028 34. [13] Hyde B, Woodside KJ. Postoperative acute respiratory distress syndrome development in the thoracic surgery patient. Semin Thorac Cardiovasc Surg 2006;18(1):28 34. [14] Jordan S, Mitchell JA, Quinlan GJ, Goldstraw P, Evans TW. The pathogenesis of lung injury following pulmonary resection. Eur Respir J 2000;15:790 9. [15] Page R, Keogh B, Kinsman R. First National Thoracic Surgery Activity and Outcomes Report 2008. The Society for Cardiothoracic Surgeons of Great Britain and Ireland. Published by Dendrite Clinical Systems Ltd., Henley on Thames, UK. [16] Slinger PD. Perioperative fluid management for thoracic surgery: the puzzle of postpneumonectomy pulmonary edema. J Cardiothorac Vasc Anesth 1995;9:442 51. [17] Zeldin R, Normandin D, Landtwing D, Peters RM. Postpneumonectomy oedema. J Thorac Cardiovasc Surg 1984;87:359 65. [18] Patel RL, Townsend ER, Fountain SW. Elective pneumonectomy: factors associated with mortality. Ann Thorac Surgery 1992;54:84 8. [19] Nohl-Oser HC. An investigation of the anatomy of the lymphatic drainage of the lungs as shown by the lymphatic spread of bronchial carcinoma. Ann R Coll Surg Engl 1972;51:157 76. [20] Gattinoni L, Mascheroni D, Torresin A, Marcolin R, Fumagalli R, Vesconi S, Rossi GP, Rossi F, Baglioni S, Bassi F, Nastri G, Pesenti A. Morphological response to positive end expiratory pressure in acute respiratory failure. Intensive Care Med 1986;12:137 42. [21] Ranieri VM, Suter PM, Tortorella, De Tullio R, Dayer JM, Brienza A, Bruno F, Slutsky AS. Effect of mechanical ventilation on inflammatory mediators in patients with acute respiratory distress syndrome: a randomized controlled trial. JAMA 1999;282:54 61. [22] Hickling KG, Walsh J, Henderson S, Jackson R. Low mortality rate in adult respiratory distress syndrome using low-volume pressure limited ventilation with permissive hypercapnia: a prospective study. Crit Care Med 1994;22:1568 78. [23] Lessard MR, Guerot E, Lorino H, Lemaire F, Brochard L. Effects of pressurecontrolled with different I:E ratios versus volume-controlled ventilation on respiratory mechanics, gas exchange and hemodynamics in patients with adult respiratory distress syndrome. Anesthesiology 1994;80:983 91. [24] Gattioni L, Tognoni G, Pesenti A, Taccone P, Mascheroni D, Labarta V, Malacrida R, Di Giulio P, Fumagalli R, Pelosi P, Brazzi L, Latini R. Effect of prone positioning on the survival of patients with acute respiratory failure. N Eng J Med 2001;345:568 73. [25] Meduri GU, Headley AS, Golden E, Carson SJ, Umberger RA, Kelso T, Tolley EA. Effect of prolonged methylprednisolone therapy in unresolving acute respiratory distress syndrome: a randomized controlled trial. JAMA 1998; 280:159 65.