Peri-operative risk factors for acute lung injury after elective ...

British Journal of Anaesthesia 86 (5): 633±8 (2001)

Peri-operative risk factors for acute lung injury after elective oesophagectomy² S. Tandon1, A. Batchelor1, R. Bullock1, A. Gascoigne1, M. Grif®n2, N. Hayes2, J. Hing3, I. Shaw1, I. Warnell1 and S. V. Baudouin1 3* 1

Departments of Anaesthesia and Intensive Care Medicine, Newcastle upon Tyne NHS Trust, Newcastle upon Tyne, UK. 2Northern Oesophago-gastric Unit, Newcastle upon Tyne NHS Trust, Newcastle upon Tyne, UK. 3University Department of Surgical and Reproductive Sciences, University of Newcastle upon Tyne, UK *Corresponding author: Department of Anaesthesia, Royal Victoria In®rmary, Newcastle upon Tyne NE1 4LP, UK Acute lung injury after oesophagectomy is well recognized but the risk factors associated with its development are poorly de®ned. We analysed retrospectively the effect of a number of pre-, peri- and post-operative risk factors on the development of lung injury in 168 patients after elective oesophagectomy performed at a single centre. The acute respiratory distress syndrome (ARDS) developed in 14.5% of patients and acute lung injury in 23.8%. Mortality in patients developing ARDS was 50% compared with 3.5% in the remainder. Features associated with the development of ARDS included a low pre-operative body mass index, a history of cigarette smoking, the experience of the surgeon, the duration of both the operation and of onelung ventilation, and the occurrence of a post-operative anastomotic leak. Peri-operative cardiorespiratory instability (measured by peri-operative hypoxaemia, hypotension, ¯uid and blood requirements and the need for inotropic support) was also associated with ARDS. Acute lung injury after elective oesophagectomy is associated with intraoperative cardiorespiratory instability. Br J Anaesth 2001; 86: 633±8 Keywords: complications, ARDS; surgery, oesophagectomy; complications, post-operative respiratory failure; complications, hypoxaemia Accepted for publication: November 11, 2000

A recent con®dential enquiry into post-operative mortality and morbidity highlighted the large contribution that respiratory complications make to poor outcome after elective oesophageal surgery.1 One speci®c pulmonary disorder, the acute respiratory distress syndrome (ARDS) is associated with oesophagectomy and occurs in 10±20% of cases.2 3 Diffuse damage to the alveolar epithelial/endothelial interface of the lung increases vascular permeability and impairs pulmonary gas exchange by alveolar ¯ooding and in®ltration with acute in¯ammatory cells. Mortality of ARDS exceeds 50%.4 The development of ARDS is associated with systemic in¯ammatory conditions such as septic shock, major trauma, massive blood transfusion and pancreatitis.5 Systemic release of circulating in¯ammatory mediators and the activation of leukocytes may contribute to the lung injury. The cause of ARDS after oesophagectomy is unknown but the release of in¯ammatory mediators and gut-related endotoxins may contribute.6 In addition, the period of one-lung anaesthesia required during oesopha-

gectomy may also be important. Collapse and subsequent re-expansion of the non-dependent lung may produce lung injury by an ischaemic/reperfusion mechanism,7 and overventilation of the dependent lung could also produce injury by micro-barotrauma.8 Recent reports of preoptimization in high-risk surgical patients suggest another mechanism of lung injury. Preoptimization techniques are used to increase cardiac output deliberately and produce so-called supranormal levels before, during and after surgery.9 A number of studies have reported decreased mortality and morbidity (including respiratory) after preoptimization.10 The corollary of these observations would be the ®nding that morbidity and mortality are increased in non-preoptimized patients who demonstrate cardiorespiratory instability in the peri-operative period. We therefore studied retrospectively the operative course of a large group of patients who underwent elective ²

This article is accompanied by Editorial II.

Ó The Board of Management and Trustees of the British Journal of Anaesthesia 2001

Tandon et al.

oesophagectomy at a single centre. We examined a number of measures of peri-operative cardiorespiratory instability and performed single and multi-variable analysis on the association between these and the development of ARDS, respiratory morbidity and mortality after elective oesophageal surgery.

Methods One hundred and ninety-three patients were operated on for oesophageal cancer at Newcastle General Hospital and the Royal Victoria In®rmary between January 1, 1996 and December 31, 1999. Of these, ®ve underwent trans-hiatal oesophagectomy, two underwent emergency subtotal oesophagectomy and 17 were found to be inoperable. One hundred and sixty-nine patients underwent conventional two-phase oesophagectomy, consisting of subtotal oesophageal resection with right thoracotomy, oesophagogastrostomy and radical two-®eld lymph node dissection. We analysed retrospectively 168 patients who underwent elective two-phase resection (the notes of one patient could not be obtained). Two patients developed severe intraoperative haemodynamic compromise that necessitated abandonment of the resection. Data were collected from patient case notes, from our prospectively compiled oesophagectomy database and from the anaesthetic charts. Missing values were recorded as such. Data were divided into preoperative, peri-operative and post-operative data. The preoperative data included age, sex, smoking history, history of ischaemic heart disease, body mass index (BMI) (kg m±2), arterial blood gases on air, simple spirometry (forced expiratory volume in one second (FEV1) and forced vital capacity (FVC) as percentage predicted and FEV1/FVC ratio), haemoglobin concentration, liver function (serum bilirubin, albumin, aspartate transaminase), renal function (blood urea and creatinine) and the stage of tumour (TNM classi®cation: stage I, IIA, IIB, III and IV). The stage of tumour was recorded from the post-operative histology. Intraoperative data included duration of surgery (from skin incision to closure), duration of one-lung ventilation (OLV), a calculated measure of the degree and duration of hypotension during the procedure (the hypotension index; see below), a calculated measure of the degree and duration of oxygen desaturation during OLV (the hypoxaemia index; see below), inotrope requirement (both bolus and infusion), estimated blood loss (ml) and blood transfusion (units), crystalloid infusion (litres), colloid infusion (litres) and the use of epidural analgesia. The experience and grade of the surgeon and anaesthetist were also recorded. All operations were performed by a consultant surgeon. The surgeon and anaesthetist were classi®ed as experienced if they had more than 5 yr of consultant experience in the management of oesophagectomies.11 Post-operative data included post-operative blood transfusion, post-operative inotrope use and the occurrence of post-operative anastomotic leak.

Post-operative complications Information on the occurrence of speci®c post-operative complications was collected. Acute lung injury (ALI) and ARDS were de®ned according to the American European Consensus Conference on ARDS criteria.12 ALI was de®ned as PaO2/FIO2 less than 40 kPa and ARDS as PaO2/ FIO2 less than 27 kPa. Additional criteria included the presence of bilateral in®ltrations on plain chest radiograph, and a pulmonary artery occlusion pressure of less than 18 mm Hg if measured or no clinical evidence of left atrial hypertension. We also de®ned a group of patients with postoperative respiratory failure on the basis of having ALI and the continuing need for invasive ventilatory support [not CPAP (continuous positive airway pressure) alone] more than 48 h after oesophagectomy. Speci®c respiratory complications included persistent pleural effusions, prolonged (>7 days) pneumothoraces and hydropneumothoraces, empyema, chylothoraces, pneumonia (fever, purulent sputum, new pulmonary in®ltrate and organisms identi®ed), pulmonary emboli, laryngeal oedema and the development of ®stulae. In-hospital deaths during the admission for oesophagectomy were also recorded.

Anaesthetic technique All of the patients in this study were anaesthetized by one of three consultant anaesthetists. When appropriate, premedication was prescribed and patients received their normal medication pre-operatively. Anaesthesia was induced i.v. and maintained with a combination of systemic opiates, neuromuscular paralysis and a volatile anaesthetic agent. Subtotal oesophagectomy is a two-phase procedure, the ®rst phase being a laparotomy conducted in the supine position and the second phase involving a right-sided thoracotomy in the left lateral position. Surgical access in the latter stage necessitates the collapse of the non-dependent lung. To facilitate a period of one-lung anaesthesia, intubation was performed using a left-sided double-lumen Robertshaw endobronchial tube. After intubation, the correct position of the double lumen tube was con®rmed by ®bre-optic bronchoscopy and patients were ventilated with an Engstrom MIE Carden Ventilator (Engstrom MIE, Exeter, Devon, UK). Initial ventilation settings were adjusted to achieve a tidal volume of 8±10 ml kg±1, a respiratory rate of 10±12 b.p.m., an inspiratory:expiratory (I:E) ratio of 1:2 and an end-tidal carbon dioxide tension of 4±5 kPa. Before the second part of the operation the patient was placed on an FIO2 of 100%, placed in the left lateral decubitus position and OLV was commenced. Ventilator settings were adjusted to maintain tidal volume and end-tidal carbon dioxide at pre-OLV levels. Peak airway pressures invariably increased because of the high resistance of the single lumen within the double-lumen tube, and no changes in ventilation were made on the basis of increased airway pressure alone. Tube displacement or malposition was always excluded by

634

ARDS after oesophagectomy

®bre-optic inspection during any period of ventilatory dif®culty. Full monitoring was established, including invasive measurement of systemic blood pressure, central venous pressure, urine output, core temperature, and the usual airway and ventilation measurements. Measures were taken to maintain normothermia. Manipulation of the non-ventilated lung to maintain systemic oxygenation included insuf¯ation of 100% oxygen (2±3 litre min±1) and the application of 5±10 cm H2O CPAP. After surgery and before transfer to the intensive care unit, the patient was reintubated with an endonasal or endotracheal tube. Sedation was maintained with i.v. propofol and opiate. Early extubation was encouraged once the patient was regarded as cardiorespiratory-stable, normothermic and comfortable and if blood loss from the chest drains was less than 50 ml h±1. When early extubation was not possible, the patient was ventilated overnight. After extubation, the patient remained in the intensive care unit until he or she could safely be transferred to a high dependency unit (HDU). Post-operative analgesia was provided by a thoracic epidural infusion of 0.125% bupivacaine containing diamorphine 40 mg ml±1, and adjusted in accordance with the patient's needs.

Area under the curve To pharmacokineticists, the area under the concentration±time curve is an important measure of exposure to a drug or metabolite. When using estimates of exposure, determination of the area under the concentration±time curve (AUC) is customarily carried out using the trapezoidal method.13 We used the analogy of hypotension or hypoxaemia as a metabolic toxin to which the patient was exposed to during the surgery. The measure of the exposure could then be determined by estimating the AUC. Hypoxaemia index

Oxygen saturation recorded during OLV was plotted on a ®xed linear scale of 0±100% against time for each patient and the AUC was calculated using the trapezoidal rule. The area was standardized for time by dividing the area by the OLV duration. Hypotension index

A plot of mean arterial pressure (MAP) against time was made and the total AUC was calculated using a MAP of less than 70 mm Hg as the cut-off value. The hypotension index was then calculated by dividing the total AUC by the total duration of surgery.

Statistical analysis Data were analysed using SPSS for Windows 8.1 (SPSS Inc., Chicago, IL, USA) under the Windows 95 operating system. Patients with data missing for a given variable were

not included in univariant analysis for that variable. Initial univariant analysis, using the independent Student t-test for discrete variables and the c2 test for categorical variables, was used to compare patient outcomes (mortality, ARDS and respiratory failure) against individual pre-, peri- and post-operative variables. Multiple logistic regression analysis, using a forward conditional strategy, was then used to examine the importance and interaction of variables with respect to the outcome.

Results ARDS occurred in 14.5% of all patients who underwent oesophagectomy for oesophageal cancer between January 1996 and December 1999. Intensive therapy unit (ITU) and hospital stays were prolonged in this group compared with patients who did not develop ARDS [mean (SD) ITU stay 16 (16) vs 1 (1) day; mean hospital stay 36 (34) vs 15 (8) days]. Mortality in the group with ARDS was 50% compared with 3.5% in the group without ARDS. ARDS-associated mortality accounted for 71% of total post-operative mortality. Post-operative respiratory failure occurred in 23.8% of all patients, and this group had a mean (SD) ITU stay of 17 (25) days. Mortality in this group was 37%. Speci®c respiratory complications occurred in 44% of all patients, including persistent pleural effusions (15%), prolonged pneumothoraces (4.1%) and hydropneumothoraces (1.8%), empyema (2.4%), chylothorax (2.4%), pneumonia (17.8%), pulmonary emboli (1.8%) and laryngeal oedema (n=1). Variables associated with the development of ARDS are shown in Table 1. Pre-operative factors included a low pre-operative BMI and a history of cigarette smoking. Perioperative instability was associated with ARDS, as indicated by the need for ¯uids and blood products during operation, the use of inotropes and peri-operative hypoxaemia and hypotension. Longer duration of procedure and length of time of OLV were also signi®cant. Postoperatively, the need for emergency re-exploration for bleeding was associated with ARDS. The experience of the consultant surgeon, but not the consultant anaesthetist, was also associated with the development of ARDS (P=0.005). Risk factors for the development of acute lung injury and for overall mortality were identical to those for ARDS. Multiple logistic regression analysis was used to determine the relative contributions of the variables to the development of ARDS (Table 2). In a forward conditional model, analysing the complete data on 146 patients, six factors were found to be signi®cant. A low BMI and a past or current history of smoking were the only pre-operative factors of signi®cance. Peri-operative instability, as indicated by hypoxaemia during OLV, hypotension and the need for ¯uids all increased the risk of post-operative ARDS. The factor with the highest odds ratio for the development of ARDS was the development of postoperative anastomotic breakdown.

635

Tandon et al. Table 1 Results of univariable analysis of pre-, peri- and post-operative factors associated with the development of ARDS after oesophagectomy. Data are mean (SE) and 95% con®dence interval (CI) or percentages. P