You are on page 1of 9

British Journal of Anaesthesia 102 (4): 50614 (2009)

doi:10.1093/bja/aep008

Advance Access publication February 17, 2009

CRITICAL CARE Does anaesthetic management affect early outcomes after lung transplant? An exploratory analysis
D. R. McIlroy1 2*, D. V. Pilcher3 and G. I. Snell4
1

Department of Anaesthesia and Perioperative Medicine, Alfred Hospital and Monash University, Melbourne, Australia. 2Department of Anesthesiology, Columbia-Presbyterian Medical Center, New York, NY, USA. 3 Department of Intensive Care Medicine and 4Department of Respiratory Medicine, Alfred Hospital, Melbourne, Australia
*Corresponding author: Department of Anesthesiology, Columbia-Presbyterian Medical Center, 622 W 168th Street, New York, NY 10032, USA. E-mail: dm2655@columbia.edu
Downloaded from http://bja.oxfordjournals.org/ by guest on October 28, 2012

Background. Primary graft dysfunction (PGD) is a predominant cause of early morbidity and mortality after lung transplantation. Although substantial work has been done to understand risk factors for PGD in terms of donor, recipient, and surgical factors, little is understood regarding the potential role of anaesthetic management variables in its development. Methods. We conducted a retrospective exploratory analysis of 107 consecutive lung transplants to determine if anaesthesia factors were associated with early graft function quantied by PaO2/FIO2. Multivariate regression techniques were used to explore the association between anaesthetic management variables and PaO2/FIO2 ratio 12 h after operation. The relationship between these variables and both time to tracheal extubation and intensive care unit (ICU) length of stay was further examined using the Cox proportional hazards. Results. On multivariate analysis, increasing volume of intraoperative colloid, comprising predominantly Gelofusinew (succinylated gelatin), was independently associated with a lower PaO2/ FIO2 12 h post-transplantation [b coefcient 242 mm Hg, 95% condence interval (CI) 27 to 277 mm Hg, P0.02] and reduced rate of extubation [hazard ratio (HR) 0.65, 95% CI 0.49 0.84, P0.001]. There was a trend for intraoperative colloid to be associated with a reduced rate of ICU discharge (HR 0.79, 95% CI 0.31 1.02, P0.07). Conclusions. We observed an inverse relationship between volume of intraoperative colloid and early lung allograft function. The association persists, despite detailed sensitivity analyses and adjustment for potential confounding variables. Further studies are required to conrm these ndings and explore potential mechanisms through which these associations may act. Br J Anaesth 2009; 102: 50614 Keywords: lung, gas exchange, respiratory; lung, transplantation Accepted for publication: January 8, 2009

Lung transplantation is the denitive treatment for endstage lung disease with 1800 transplants per year performed worldwide.1 Despite 1 yr survival approaching 80%, primary graft dysfunction (PGD) remains a predominant cause of early morbidity and mortality. PGD is variably reported with an incidence of 10 50% and a 30 or 90 day mortality of 14 63% of those affected,2 5 the variation due in part to inconsistencies in denitions. The International Society of Heart and Lung Transplantation

(ISHLT) published recommendations on the classication and nomenclature to be used in PGD6 with severity quantied by the measured ratio of PaO2 to FIO2 in combination with the presence or absence of inltrates on the chest X-ray (CXR) from 0 to 72 h after operation. However, in view of inherent inconsistencies in CXR reporting, a modied ISHLT grading system has been subsequently described and validated,4 based solely on the observed PaO2/FIO2 ratio (Table 1).

# The Author [2009]. Published by Oxford University Press on behalf of The Board of Directors of the British Journal of Anaesthesia. All rights reserved. For Permissions, please email: journals.permissions@oxfordjournal.org

Anaesthesia and lung transplant outcomes

Table 1 Modied ISHLT recommendations for grading of PGD severity, based solely on measured PaO2/FIO2 ratio. [The PaO2/FIO2 ratio is calculated as the ratio of partial pressure of oxygen in arterial blood (mm Hg) to fractional inspired oxygen concentration (1.0)] PGD grade 1 2 3 PaO2/FIO2 .300 200 300 ,200

and donor organ ischaemic time. For bilateral transplants, the ischaemic time was taken as the ischaemic time of the second lung. Intraoperative variables Operation (bilateral vs single), duration of surgery, epidural catheterization, antibrinolytic, inhaled nitric oxide or cardiopulmonary bypass (CPB) use, total volume of i.v. uid administered including volume of individual uid components, central venous (CVP), and pulmonary artery pressures (PAS/PAD), inotrope and vasodilator use, and, in bilateral transplantation, time at which the entire cardiac output was directed through the rst implanted lung were all recorded. CVP-1 was dened as CVP immediately before induction of anaesthesia. Where not available, the rst recorded CVP after induction of anaesthesia was used. CVP-2 was dened as the rst recorded CVP in the intensive care unit (ICU). Where not available, the last recorded CVP before departing the operating room (OR) was used. The same denitions applied to recording pulmonary artery pressures. Vasoactive medication use was recorded as peak sustained dose (.30 min) during surgery. Data were recorded for epinephrine, norepinephrine, and sodium nitroprusside. Biochemical data during surgery were recorded, including highest and lowest measured haemoglobin concentration, highest recorded PaCO2, lowest recorded arterial pH, and highest recorded blood glucose level. Postoperative variables A new management protocol to standardize postoperative cardiorespiratory and uid management of these patients17 was introduced in September 2005 and was included for analysis. Outcome variables PaO2/FIO2 ratio was recorded at time 0, 3, 6, 12, 24, 48, and 72 h post-transplantation and time to extubation, ICU LOS, and 30 day mortality.

PGD presents a wide spectrum of disease severity characterized by varying degrees of impairment of gas exchange and is associated with delayed extubation, prolonged intensive care and hospital length of stay (LOS), increased early mortality, and worse long-term outcome among survivors.7 11 Treatment is essentially supportive. Numerous studies have attempted to dene factors associated with increased risk of PGD. Most published work in this area has focused on donor variables, preoperative recipient variables, surgical factors, and early postoperative management.5 12 16 However, there is a current gap in the literature assessing the impact of anaesthetic management variables on early graft function after lung transplantation. The aim of the current study is to explore the relationship between such factors and early lung allograft function. We hypothesize that anaesthetic management factors will account for signicant postoperative outcome variability, providing potential targets for therapeutic strategies to optimize early graft function.

Downloaded from http://bja.oxfordjournals.org/ by guest on October 28, 2012

Methods
Study group
We retrospectively studied 107 consecutive patients undergoing lung transplantation at The Alfred Hospital between May 2004 and January 2007. This group comprised 84 bilateral transplants (two re-transplants) and 23 single transplants. Combined heart lung transplants were excluded from analysis.

Donor and recipient matching Data extraction and denitions


After IRB approval, data were retrospectively extracted from the medical record by two trained reviewers. Preoperative recipient variables Baseline patient characteristics were recorded and also the patients underlying lung pathology, classied as obstructive, restrictive, or vascular [including primary pulmonary hypertension (PPH)]. Obstructive pathology was further classied as cystic brosis (CF) or non-CF related. Donor variables These included age, gender, donor PaO2 at retrieval, inotrope use at organ procurement, preservation solution used, Donor assessment and management and donor recipient matching proceeded according to standard protocols described elsewhere.18 19 Although donor selection is based around standard criteria, marginal donors are commonly considered and used within our institution. Before operation, a panel-reactive antibody assay was used to screen for preformed antibodies to human leucocyte antigens and prospective donor recipient T, and B-cell cross-matching was performed.

Lung procurement and preservation


Lung procurement and preservation followed standard procedures.20 21 Attempts were made to optimize donor haemodynamics with i.v. uid titrated to a CVP of 10 mm Hg

507

McIlroy et al.

with subsequent titration of norepinephrine to maintain a suitable mean arterial pressure until aortic cross-clamp application. Before September 2004, an i.v. infusion of prostacyclin 40 80 ng kg21 min21 was administered for 10 min before a single antegrade ush with cold modied Euro-Collins solution (60 ml kg21). After September 2004, Perfadex (Vitrolife, Goteborg, Sweden) replaced the combination of prostacyclin and Euro-Collins at our institution. Organs procured by other institutions and forwarded to us were preserved using Papworth solution. Donor lungs were inated with an FIO2 1.0 before tracheal stapling.

Postoperative management
Before September 2005, postoperative management in the ICU aimed at ensuring adequate end-organ perfusion. Although low lling pressures were targeted, they were not aggressively pursued.13 From September 2005, a new postoperative uid management regimen was instituted in our lung transplant recipients, which has been previously described.17 This encompassed respiratory and cardiovascular management algorithms, targeting a CVP ,7 mm Hg where mean arterial pressure and cardiac index permitted. The protocol provided an algorithm for early extubation where the PaO2/FIO2 ratio was .200. Inhaled nitric oxide was used (20 ppm) if there was increasing pulmonary artery pressures or decreasing oxygenation and extracorporeal membrane oxygenation if required. Immunosuppression was achieved as per protocols previously published.18 All donors and recipients received methylprednisolone (1 g) intraoperatively. Recipients also received immunosuppressive induction with cyclosporine and azathioprine at the time of surgery. All patients received prophylactic antibiotics on the basis of known or suspected donor and recipient microbiology results and ganciclovir for cytomegalovirus prophylaxis where indicated.

Operative procedure
Premedication was typically not used. All patients were anaesthetized using a standardized balanced total i.v. technique consisting of fentanyl (15 25 mg kg21), midazolam (0.05 0.1 mg kg21), and propofol by continuous infusion (50 100 mg kg21 min21), titrated to maintain a bispectral index value ,60. Intraoperative haemodynamic monitoring consisted of invasive arterial pressure monitoring, pulmonary artery catheter, and transoesophageal echocardiography (TOE). Intraoperative uid management was at the discretion of the treating anaesthetist, Gelofusinew (B. Braun, Australia) being used for colloid therapy and either lactated Ringers or normal saline for crystalloid replacement therapy. With the exception of PPH, CPB was not routinely used to facilitate the operative procedure. Where required, CPB was instituted in the setting of severe haemodynamic instability or severe refractory hypoxaemia and was accomplished with a roller-pump-driven continuous ow device (Jostra, Maquet, Hirrlingen, Germany) via an uncoated circuit. A 4% albumin solution, 500 ml (Albumexw 4, CSL, Broadmeadows, Australia), was added to the circuit prime, representing the only other colloid solution used in this series. Body temperature was maintained at 35 36.58C. Before completion of implantation, retrograde followed by antegrade reperfusion and de-airing was performed through an untied pulmonary arterial anastomotic suture line. A specic pressure and ow-controlled reperfusion technique was not used. After reperfusion of transplanted lungs, ventilation commenced with an FIO2 of 1.0, PEEP 5 cm H2O21, and an I:E ratio of 1:2, aiming to maintain peak airway pressures ,25 cm H2O21. Ventilatory frequency was adjusted to achieve the desired PaCO2 and FIO2 reduced to 0.5 0.8 as tolerated, while maintaining the SpO294%. TOE facilitated routine intraoperative examination of pulmonary venous anastomoses before leaving the OR.22 Although sited before operation, epidural catheters were only used once both transplanted lungs were reperfused and the patient was haemodynamically stable. After a loading dose of fentanyl 100 mg, a maintenance infusion of ropivacaine 2 mg ml21 and fentanyl 2 mg ml21 was commenced at 6 10 ml h21.

Downloaded from http://bja.oxfordjournals.org/ by guest on October 28, 2012

Statistical analysis
Statistical analysis was performed with Stata version 10 (StataCorp, College Station, TX, USA). Continuous data were assessed for normality and expressed as mean (SD) or median and inter-quartile range. Categorical data are expressed as count and proportions. The primary outcome was to determine if anaesthesia management factors were signicantly associated with early postoperative allograft function measured by the PaO2/FIO2 ratio. Using the Cox proportional hazards, an initial assessment was made of the relationship between PaO2/FIO2 at each measured time point (0 72 h after operation) with the outcomes of both time to extubation and ICU LOS. The time point with the strongest statistical association with clinical outcome was selected as the dependent variable for further exploratory analysis with intraoperative data. After selecting the time point for assessment of graft function (PaO2/FIO2), univariate regression analysis with each of the explanatory variables in our data set was performed, including donor lung preservation protocol and postoperative uid management regimen and also factors previously demonstrated to be associated with early graft function for potential inclusion in a multivariate model.2 5 7 12 14 Variables were assessed for normality and mathematically transformed where appropriate. Factors signicant to a P-value of 0.2 on univariate analysis were incorporated into a multivariate model. Manual backward stepwise elimination was performed, excluding the most non-signicant variable at each step. This process continued until there were at least 10 outcomes for each

508

Anaesthesia and lung transplant outcomes

Table 2 Distribution of donor and recipient variables (n107). [Continuous variables are given as mean (range) or mean (SD), categorical variables as count (proportion)] Donor variable Donor age (yr) Donor sex (male) Donor PaO2 at procurement (mm Hg) Donor inotropic support (at procurement) Lung ischaemic time (min) Preservation solution Euro-Collins (,September 2004) Perfadex (.September 2004) Papworth Recipient variable Age (yr) Recipient sex (male) Weight (kg) Height (m) Body surface area (BSA) (m2) Body mass index (BMI) (kg m22) Recipient lung pathology Obstructive (cystic brosis) Obstructive (non-cystic brosis) Restrictive Vascular 38 (11 64) 60 (57.1%) 442 (95) 88 (93.6%) 349 (131) 8 (7.5%) 79 (73.8%) 20 (18.7%) 47 (16 65) 53 (49.5%) 63 (14) 1.67 (0.10) 1.70 (0.22) 22.2 (4.1) 24 56 21 6 (22.4%) (52.3%) (19.6%) (5.6%)

Table 3 Distribution of intraoperative recipient variables (n107). [Continuous variables are given as mean (SD), categorical variables as count (proportion).] *CVP-1 or PA pressure-1 refers to the pressure immediately before induction of general anaesthesia. When this was not available, the earliest recorded pressure post-induction of anaesthesia was used. CVP-2 or PA pressure-2 refers to the rst recorded pressure in the ICU. Where this was not available, the last recorded pressure before departing the OR was used Intraoperative variable Operation Single lung transplant Bilateral lung transplant Duration of surgery (h) Epidural catheterization in OR Yes No Paravertebral Antibrinolytic None Aprotinin Tranexamic acid Inhaled nitric oxide Lowest intraoperative haemoglobin (g litre21) Highest intraoperative haemoglobin (g litre21) Highest intraoperative PaCO2 (mm Hg) Lowest intraoperative pH Highest blood glucose (g dl21) Lowest intraoperative temperature (8C) Total i.v. uid volume intraoperatively (litre) Crystalloid volume (litre) Colloid volume (litre) Volume of cell-saved (autologous) blood (ml) Packed red blood cells (ml) Fresh-frozen plasma/cryoprecipitate (ml) Platelets (ml) No. of patients who received 1 unit packed red cells No. of patients who received FFP/cryoprecipitate No. of patients who received 1 pool of platelets No. of patients who received any blood products Cardiopulmonary bypass used CVP-1* (mm Hg) CVP-2 (mm Hg) PA pressure-1* (mm Hg) PA pressure-2 (mm Hg) Time for entire pulmonary blood ow through rst implanted lung (where time .0) (min)

23 (21.5%) 84 (78.5%) 5.1 (1.7) 82 (77.4%) 21 (19.8%) 3 (2.8%) 43 (40.2%) 61 (57%) 3 (2.8%) 38 (35.9%) 96.7 (22.1) 142.6 (20.7) 75.2 (25.0) 7.22 (0.11) 12.6 (3.3) 35.2 (0.8) 4.64 (2.27) 2.74 (1.34) 1.11 (0.96) 294 (450) 283 (428) 191 (353) 33 (110) 44 (41.1%) 31 (29.0%) 10 (9.4%) 57 (53.3%) 19 (17.8%) 9.9 (4.8) 6.6 (4.0) 44.8 (21.2)/22.9 (9.8) 28.5 (8.8)/14.7 (5.7) 65 (31)

Downloaded from http://bja.oxfordjournals.org/ by guest on October 28, 2012

variable remaining in the model, further elimination proceeding only if it produced a more parsimonious model. Variables that remained are expressed as regression coefcient, with P-value and 95% condence interval (CI). A multivariate Cox regression model was then used to assess the relationship between the variables retained in the linear regression model and clinical outcomes of both time to extubation and ICU LOS. These results are expressed as hazard ratio (HR) with P-value and 95% CI. Further sensitivity analyses were performed to assess the effect of specic subgroups or adding potential confounding variables to the results.

Results
Baseline characteristics of donors and recipients along with donor operative data are provided in Table 2. Recipient intraoperative variables are detailed in Table 3. Median time to extubation was 0.9 days (inter-quartile range 0.5 2.2), median ICU LOS was 3.2 days (interquartile range 2 6), and 30 day mortality was 2%. Forty-three subjects (40%) had at least one PaO2/FIO2 ,200 recorded within the rst 6 postoperative hours. However, by 12 h after operation, only 13 subjects (12%) had a PaO2/ FIO2 ratio ,200, this number continuing to decline through the remainder of the measurement period (Fig. 1). Of all time points examined from 0 to 72 h, the PaO2/FIO2 ratio 12 h post-transplantation (T12) demonstrated the strongest statistical association with both time to extubation (HR 1.003, P0.001) and ICU LOS (HR 1.005, P,0.001). On univariate regression CVP-2 (P0.03), PA-systolic pressure-2 (P0.02), use of CPB (P0.05), single vs bilateral transplantation (P0.02), and epidural catheterization (P0.01) were the only intraoperative variables

signicantly associated with PaO2/FIO2 (T12) (Table 4), although there was a trend for increased volume of i.v. colloid to be associated with a reduced PaO2/FIO2 (T12) (P0.06). Other donor and recipient variables previously demonstrated to be associated with early graft function were also assessed for potential inclusion in the multivariate model. There was no association between use of the postoperative uid management algorithm and PaO2/ FIO2 ratio 12 h post-transplantation. A multivariate regression model was constructed for PaO2/FIO2 (T12) according to the methods described. Where PaO2/FIO2 (T12) was unavailable, imputation with PaO2/FIO2 (T24) was performed (n12). The initial model contained 12 explanatory variables. Colloid volume underwent loge transformation to better approximate a linear relationship. However, 18 subjects received no intraoperative colloid making transformation of their data impossible. Combined with missing data, the number of subjects available for

509

McIlroy et al.

PaO2/FIO2 ratio, 072 h post-transplantation 120 100 Cohort (%) 80 60 40 20 0 0 3


15 23 33 36 26 20 13 13 5 50

43

36 39 25

PaO2/FIO2>300 PaO2/FIO2 200300 PaO2/FIO2<200

12 14 4 6 9 1

23

6 12 24 Time after transplantation (h)

48

72

Fig 1 Distribution of patients by assigned PGD grade at each time point measured through 72 h after operation (horizontal axis). The vertical axis represents the proportion of the entire cohort within each PGD grade; absolute numbers are on the histogram bars. The decreasing number of patients with an assigned PGD grade, most marked beyond 12 h, represents the increasing proportion of patients achieving tracheal extubation, removal of invasive arterial monitoring line, and discharge from the ICU.

Downloaded from http://bja.oxfordjournals.org/ by guest on October 28, 2012

analysis was reduced to 68. Stepwise elimination proceeded until four variables remained. Likelihood testing conrmed this to be the most parsimonious model (Table 5). Intraoperative colloid administration (b coefcient 242 mm Hg, 95% CI 27 to 277 mm Hg, P0.02) and recipient pathology of CF (b coefcient 52 mm Hg, 95% CI 8.3 109 mm Hg, P0.02) were associated with the outcome. The loge transformation of intraoperative colloid indicates that the PaO2/FIO2 (T12) is reduced by 42 for every 2.7-fold increase in the volume of colloid administered. With an adjusted R 2 value of 0.21, this multivariate model accounts for more than one-fth of the observed variation in PaO2/FIO2 (T12). The model conformed to the assumptions of linear regression with no evidence of collinearity between variables. The effect of adding other intraoperative uid and blood product administration variables to the model was explored, but none of these achieved signicance nor was overall t improved. Sequential manual addition of other potential confounders to the model, including use of CPB, epidural catheterization, CVP-2, surgical duration, or use of the postoperative ICU uid management regimen, did not materially alter results. For further sensitivity analysis, patients with a pulmonary vascular aetiology for their underlying lung disease were excluded with no signicant change in results of multivariate analysis. Further exclusion of patients requiring CPB support to facilitate surgery resulted in i.v. colloid administration failing to achieve signicance albeit with a minimal change to the regression coefcient (b coefcient 236 mm Hg, 95% CI 284 to 12 mm Hg, P0.14). Repeating the analysis by each class of recipient pathology, the inverse association between volume of colloid and PaO2/FIO2 (T12) persisted (b-regression coefcient range: 231 to 2105 mm Hg), although with reduced numbers in each group, it was no longer statistically signicant. Cox regression techniques were used to assess the relationship between the variables derived from the

multivariate linear regression model and the outcomes of time to extubation and ICU LOS (Table 6). Intraoperative colloid was signicantly associated with a reduced rate of extubation (HR 0.65, 95% CI 0.49 0.84, P0.001) as was a pulmonary vascular aetiology for the primary lung disease (HR 0.25, 95% CI 0.07 0.90, P0.03). Intraoperative colloid remained signicantly associated with a reduced rate of extubation, despite excluding cases performed with the aid of CPB (HR 0.63, 95% CI 0.46 0.87, P0.01) or cases performed without the use of an epidural (HR 0.66, 95% CI 0.48 0.90, P0.01). Repetition of the analysis by pre-transplant recipient lung pathology demonstrated a consistent albeit non-signicant trend towards a reduced HR for time to extubation across each group. There was a trend for increasing volume of i.v. colloid to be associated with a reduced rate of ICU discharge (HR 0.79, 95% CI 0.31 1.02, P0.07) whereas increasing PA-systolic pressure-2 was associated with a reduced rate of ICU discharge (HR 0.96, 95% CI 0.93 0.99, P0.01). However, adjusting the analysis for either epidural use or CVP-2 rendered the association between PA pressure and rate of ICU discharge non-signicant. Further repetition of the analysis according to pretransplant lung pathology demonstrated a consistent trend across groups towards a reduced rate of ICU discharge with increasing intraoperative colloid. Colloid volume was included in the Cox model without the need for loge transformation and model checking conrmed that neither model violated the assumptions of proportional hazards.

Discussion
Previous studies investigating the association between intraoperative variables and PGD have focused on single interventions or measures, such as the use of inhaled nitric oxide, aprotinin, or CPB, different preservation solutions, and ushing techniques of the lungs and ischaemic times.21 23 26 In the current study, multivariate regression

510

Anaesthesia and lung transplant outcomes

Table 4 Univariate associations between intraoperative, donor and recipient variables, and PaO2/FIO2 (T12) (n107). [PaO2/FIO2 (T12) is the PaO2/FIO2 at 12 h post-transplant.] *Variable has undergone loge transformation. CVP-1 or PA pressure-1 refers to the pressure immediately before induction of general anaesthesia. When this was not available, the earliest recorded pressure post-induction of anaesthesia was used. CVP-2 or PA pressure-2 refers to the rst recorded pressure in the ICU. Where this was not available, the last recorded pressure before departing the OR was used. PRBC, packed red blood cells; FFP, fresh-frozen plasma. BSA, body surface area; BMI, body mass index; CF, cystic brosis; EC, Euro-Collins solution PaO2/FIO2 (T12) b regression coefcient Intraoperative variable Colloid volume* Crystalloid volume Total uid volume* PRBCs used FFP/cryoprecipitate used Platelets used Cardiopulmonary bypass used Epidural sited in OR Aprotinin used Inhaled nitric oxide used CVP-1 CVP-2 PA systolic-1 PA systolic-2 Peak PaCO2 (mm Hg)* Peak blood glucose (g litre21)* Lowest temperature (8C) Time for entire CO through rst implanted lung (min)* Operation (single vs bilateral) referencebilateral Duration of surgery (h) Recipient and donor variables Recipient age Recipient (female) sex Recipient BSA (m2) Recipient BMI (kg m22) Recipient path (refnon-CF obstructive) Obstructive (CF) Restrictive Vascular Donor age .40 yr Donor (female) sex Donor PaO2 (mm Hg) Preservation solution (refEC) Perfadex Papworth Lung ischaemic time (min) Standardized postoperative ICU uid regimen (used after September 2005) P-value

Table 5 Multivariate association with PaO2/FIO2 (T12). [PaO2/FIO2 (T12) is the PaO2/FIO2 at 12 h post-transplant.] Adjusted R 20.21 (n68). *Colloid volume is loge transformed. The regression coefcient represents the change in PaO2/ FIO2 (T12) for a 2.7-fold increase in administered volume. PA systolic pressure-2 refers to the rst recorded pressure in the ICU. Where this was not available, the last recorded pressure before departing the OR was used. CF, cystic brosis Variable Regression coefcient 242 22.5 95% CI (mm Hg) 27 to 277 24.7 to 0.8 P-value

239.1 10.5 26.2 220.3 14.2 223.8 257.4 75.3 28.3 215.6 22.3 26.7 230.3 23.0 22.3 56.4 219.6 236.3 266.8 25.4 21.2 16.8 227.2 23.0 38.9 233.6 262.2 235.9 33.3 0.27 20.9 1.9 20.0002 25.8

0.06 0.28 0.33 0.39 0.57 0.54 0.05 0.01 0.24 0.52 0.32 0.03 0.33 0.02 0.96 0.22 0.22 0.43 0.02 0.41 0.15 0.47 0.62 0.31 0.16 0.29 0.24 0.12 0.15 0.03 0.70 0.98 0.06 0.81

Colloid volume* PA systolic pressure-2 Recipient pathology (refobstructive, non-CF) Obstructive (CF) Restrictive Vascular Donor PaO2 (mm Hg)

0.02 0.16

58 217 273 0.2

8.3 109 285 to 51 2178 to 32 20.02 to 0.4

0.02 0.63 0.17 0.08

Downloaded from http://bja.oxfordjournals.org/ by guest on October 28, 2012

Table 6 Multivariate Cox proportional hazards for time to extubation (n90) and ICU discharge (n90). *PA systolic pressure-2 refers to the rst recorded systolic pulmonary artery pressure in the ICU. Where this was not available, the last recorded pressure before departing the OR was used. CF, cystic brosis Variable Hazard ratio 0.65 0.99 95% CI P-value

Colloid volume (litre) PA systolic pressure-2* Recipient pathology (refobstructive, non-CF) Obstructive (CF) Restrictive Vascular Donor PaO2 (mm Hg) ICU discharge Colloid volume (litre) PA systolic pressure-2* Recipient pathology (refobstructive, non-CF) Obstructive (CF) Restrictive Vascular Donor PaO2 (mm Hg)

0.49 0.84 0.96 1.02

0.001 0.50

1.20 0.99 0.25 0.99 0.79 0.96

0.67 2.14 0.51 1.93 0.07 0.90 0.99 1.00 0.31 1.02 0.93 0.99

0.54 0.98 0.03 0.47 0.07 0.01

3.31 2.94 0.49 0.99

1.80 6.1 1.51 5.74 0.15 1.64 0.99 1.00

0.001 0.002 0.25 0.91

analysis is used to explore the association between a large number of intraoperative and anaesthesia management variables and early allograft function. We found an independent inverse association between volume of intraoperative colloid, comprising predominantly gelatin, and PaO2/FIO2 12 h post-transplantation. Furthermore, each additional litre of intraoperative colloid was associated with a 35% reduction in the rate of extubation (HR 0.65, 95% CI 0.490.84, P0.001) and a trend towards a reduced rate of ICU discharge (HR 0.79, 95% CI 0.311.02, P0.07). Although the ISHLT PGD guideline recommends assessment of PaO2/FIO2 at the time points of 0, 24, 48, and

72 h post-transplant,6 Oto and colleagues27 have demonstrated a signicant variation in the ratio within the rst 12 postoperative hours, stabilizing thereafter. They further note the early predictive value of PaO2/FIO2 (T12) for the clinical outcomes of time to extubation and ICU LOS and our data are consistent with this nding. Furthermore, the T12 time point may represent an ideal time for assessing the impact of intraoperative interventions, allowing time for optimal ventilation to be established in the ICU and before the impact of later interventions become more prominent. The stabilization of PaO2/FIO2 beyond 12 h justies the imputation of PaO2/FIO2 (T24) data where T12 data were unavailable. It is important to consider potential sources of confounding. Elevated CVP has been associated with early graft dysfunction and prolonged time to extubation.13 However, there was no association between volume of colloid

511

McIlroy et al.

administered and CVP-2 (r0.03) in the current study cohort, nor did adjusting for CVP-2 alter the association between colloid and PaO2/FIO2 (T12). Furthermore, there was no association between colloid administration and either inotrope use or pulmonary artery diastolic pressure supporting an effect of colloid that is independent of cardiac and haemodynamic performance. CPB is another potential confounder with the altered capillary permeability and inammatory response that it may induce along with the obligatory volume of colloid (4% albumin) these patients receive from the circuit prime.28 Excluding these patients from analysis resulted in minimal change to the observed association between PaO2/FIO2 (T12) and volume of intraoperative colloid (Gelofusinew) (b coefcient 250 mm Hg, 95% CI 2106 to 5 mm Hg, P0.07), although the reduction in number of subjects for analysis results in a widening of condence limits and loss of statistical signicance, making interpretation difcult. Although a number of studies have attempted to dene the association between CPB and early graft dysfunction, they have produced conicting results and the role of CPB as an independent risk factor for PGD remains controversial.16 Intraoperative colloid was associated with a 35% reduction in the rate of extubation for every additional litre administered. This relationship persisted, despite a detailed sensitivity analysis with both exclusion of potentially confounding subgroups and manual addition of potential confounding variables to the model, supporting the clinical relevance of the observed inverse association between colloid and PaO2/ FIO2 (T12). A vascular aetiology for the underlying lung disease (PPH in our population) was also signicantly associated with a reduced rate of extubation. This has been previously documented as a risk factor for PGD, supporting the consistency of the current data set with previous reports. PGD is characterized by diffuse alveolar damage29 30 and increased capillary permeability,31 32 and the recent ISHLT working group report makes general recommendation to avoid excessive i.v. uid in postoperative management of these patients.33 However, few previous studies have explored the relationship between perioperative uid therapy and early graft function. Pilcher and colleagues14 found no association between postoperative uid balance and PaO2/FIO2 ratios in the rst 24 h after operation, but they did not assess the effect of specic uid components. Although we also found no association between total i.v. uid volume, crystalloid or blood products given intraoperatively, and early graft function, the nding of a highly signicant inverse association between volume of intraoperative colloid and PaO2/FIO2 (T12) requires further consideration. One potential explanation for this nding may be the increased capillary permeability occurring with PGD allowing leakage of relatively large molecules of colloid solution into the extravascular lung space. The complete absence of lymphatic drainage in the transplanted lung may further slow the removal of these oncotically active molecules from the lung. Although this may seem

counter-intuitive given the physico-chemical properties of colloid solutions for maintaining intravascular oncotic pressure,34 the leakage of radio-labelled transferrin (molecular weight 75 000 Da) and other proteins into extravascular lung tissue is well described in patients with acute respiratory distress syndrome.35 36 Gelofusinew was the predominant intraoperative colloid in this series of patients, accounting for more than 90% of administered colloid volume. It is a succinylated gelatin, molecular weight 30 000 Da, which undergoes minimal metabolism and is excreted largely unchanged in the urine. Although a 4% albumin solution was added to the CPB priming volume as described, this represented only 7.5% of total colloid volume administered, always in the context of the CPB prime. Therefore, separate analysis was not performed. It is unclear if our ndings may be generalized to other colloid solutions. Although the introduction of a postoperative uid management algorithm was not associated with early graft function, the data were unable to assess the impact of the algorithm on specic component uid volumes administered through the early postoperative period, and this may have provided further useful information. Increasing pulmonary artery systolic pressure at end-operation was associated with a reduced rate of ICU discharge, although the association was lost when the use of CPB was adjusted for or when such cases were excluded from the analysis. Previous studies assessing the relationship between pulmonary artery pressures and early graft function have focused on elevations in preoperative pulmonary artery pressure, the signicance of which remains uncertain.5 16 Although postoperative elevations in pulmonary artery pressure may indicate serious problems with pulmonary vascular anastomoses, there was no evidence of this in the current series. Furthermore, the absence of an association between pulmonary artery pressures and either PaO2/FIO2 (T12) or time to extubation makes this association difcult to interpret, raising the likelihood that it represents a spurious nding. The lack of association between many of the other variables analysed, including those that may be considered issues of anaesthesia management, and measures of early graft function should not be interpreted to mean that their diligent control is unnecessary. There are two other conservative and more likely interpretations of this nding. The rst, a classic type II error, is the possibility of missing signicant associations in a data set of limited size. The second plausible explanation is that within the bounds of which these many variables were maintained, they did not exert an observable effect on the outcome. However, our data make no inference as to the potential effect of these variables outside the limits that we observed.

Downloaded from http://bja.oxfordjournals.org/ by guest on October 28, 2012

Study limitations
Despite a detailed sensitivity analysis, the possibility of unrecognized or unmeasured confounding cannot be

512

Anaesthesia and lung transplant outcomes

eliminated in a retrospective study and our results should be interpreted with caution. Although we describe a number of positive ndings, regression methods can only show association. These results are best used to generate hypotheses for subsequent testing under appropriate conditions. PaO2/FIO2 was chosen as the primary outcome in this study as it is an accepted measure of early graft function in lung transplantation and forms the basis for the quantication of PGD. Furthermore, in an exploratory study, the continuous nature of this outcome optimizes the possibility of detecting signicant associations that may otherwise be missed with a binary outcome such as severe PGD. It is a non-specic outcome and may represent pathology other than PGD, such as atelectasis, sputum plugging, pneumothorax, or other mechanical vascular pathologies, although the association between intraoperative colloid and increased time to extubation supports the potential clinical relevance of this nding. The 12 h time point for measurement of PaO2/FIO2 was selected for further analysis in this study on the basis of demonstrated association with duration of mechanical ventilation and ICU LOS. Despite the importance of these short-term outcomes, the current study is unable to make any inference of association between colloid volume and later outcomes. Multiple analyses have been conducted enhancing the possibility of a type I error, although the consistently adverse association of intraoperative colloid with multiple outcomes lessens that likelihood. The sample size of 107 cases is small compared with the number of potential explanatory variables considered, limiting power, and increasing potential for type II errors. Our ndings would be further strengthened if reproduced by other clinical units.

References
1 Trulock EP, Edwards LB, Taylor DO, et al. Registry of the International Society for Heart and Lung Transplantation: twenty-third ofcial adult lung and heart lung transplantation report2006. J Heart Lung Transplant 2006; 25: 880 92 2 Christie JD, Kotloff RM, Pochettino A, et al. Clinical risk factors for primary graft failure following lung transplantation. Chest 2003; 124: 1232 41 3 Meyers BF, de la Morena M, Sweet SC, et al. Primary graft dysfunction and other selected complications of lung transplantation: a single-center experience of 983 patients. J Thorac Cardiovasc Surg 2005; 129: 1421 9 4 Prekker ME, Nath DS, Walker AR, et al. Validation of the proposed International Society for Heart and Lung Transplantation grading system for primary graft dysfunction after lung transplantation. J Heart Lung Transplant 2006; 25: 371 8 5 Whitson BA, Nath DS, Johnson AC, et al. Risk factors for primary graft dysfunction after lung transplantation. J Thorac Cardiovasc Surg 2006; 131: 73 80 6 Christie JD, Carby M, Bag R, et al. Report of the ISHLT working group on primary lung graft dysfunction part II: denition. A consensus statement of the International Society for Heart and Lung Transplantation. J Heart Lung Transplant 2005; 24: 1454 9 7 Christie JD, Van Raemdonck D, de Perrot M, et al. Report of the ISHLTworking group on primary lung graft dysfunction part I: introduction and methods. J Heart Lung Transplant 2005; 24: 14513 8 Christie JD, Kotloff RM, Ahya VN, et al. The effect of primary graft dysfunction on survival after lung transplantation. Am J Respir Crit Care Med 2005; 171: 1312 6 9 Fiser SM, Tribble CG, Long SM, et al. Ischemia reperfusion injury after lung transplantation increases risk of late bronchiolitis obliterans syndrome. Ann Thorac Surg 2002; 73: 1041 7 10 King RC, Binns OA, Rodriguez F, et al. Reperfusion injury signicantly impacts clinical outcome after pulmonary transplantation. Ann Thorac Surg 2000; 69: 1681 5 11 Thabut G, Vinatier I, Stern J-B, et al. Primary graft failure following lung transplantation: predictive factors of mortality. Chest 2002; 121: 1876 82 12 de Perrot M, Bonser RS, Dark J, et al. Report of the ISHLT working group on primary lung graft dysfunction part III: donorrelated risk factors and markers. J Heart Lung Transplant 2005; 24: 1460 7 13 Pilcher DV, Scheinkestel CD, Snell GI, et al. High central venous pressure is associated with prolonged mechanical ventilation and increased mortality after lung transplantation. J Thorac Cardiovasc Surg 2005; 129: 912 8 14 Pilcher DV, Snell GI, Scheinkestel CD, et al. High donor age, low donor oxygenation, and high recipient inotrope requirements predict early graft dysfunction in lung transplant recipients. J Heart Lung Transplant 2005; 24: 1814 20 15 Christie JD, Kotloff RM, Pochettino A, et al. Clinical risk factors for primary graft failure following lung transplantation. Chest 2003; 124: 1232 41 16 Barr ML, Kawut SM, Whelan TP, et al. Report of the ISHLT working group on primary lung graft dysfunction part IV: recipient-related risk factors and markers. J Heart Lung Transplant 2005; 24: 1468 82 17 Currey JA, Pilcher DV, Davies A, et al. A prospective study of a management algorithm to minimise lung transplant (LTx) primary graft dysfunction (PGD). J Heart Lung Transplant 2007; 26: S66 7 18 Gabbay E, Williams TJ, Grifths AP, et al. Maximizing the utilization of donor organs offered for lung transplantation. Am J Respir Crit Care Med 1999; 160: 265 71

Downloaded from http://bja.oxfordjournals.org/ by guest on October 28, 2012

Conclusion
We have performed a retrospective analysis of factors associated with early graft function in lung transplantation, focusing on anaesthetic management factors. We have demonstrated an inverse association between volume of intraoperative colloid, comprising predominantly gelatin, and PaO2/FIO2 ratio 12 h after operation, and also a reduced rate of extubation. These associations persist, despite adjusting for potential confounding factors, providing potential hypotheses for subsequent testing. Further studies are required to conrm these ndings and to explore potential mechanisms through which these associations may act.

Funding
This work was supported by intramural funding from the Departments of Anaesthesia and Perioperative Medicine, Intensive Care Medicine, and Respiratory Medicine at The Alfred Hospital.

513

McIlroy et al.

19 Oto T, Grifths AP Levvey B, et al. A donor history of smoking , affects early but not late outcome in lung transplantation. Transplantation 2004; 78: 599 606 20 de Perrot M, Keshavjee S. Lung preservation. Semin Thorac Cardiovasc Surg 2004; 16: 300 8 21 Oto T, Rabinov M, Grifths AP, et al. Unexpected donor pulmonary embolism affects early outcomes after lung transplantation: a major mechanism of primary graft failure? J Thorac Cardiovasc Surg 2005; 130: 1146 52 22 McIlroy DR, Cairo SA, Buckland MR. Pulmonary vein thrombosis, lung transplantation, and intraoperative transesophageal echocardiography. J Cardiothorac Vasc Anesth 2006; 20: 712 5 23 Bittner HB, Richter M, Kuntze T, et al. Aprotinin decreases reperfusion injury and allograft dysfunction in clinical lung transplantation. Eur J Cardiothorac Surg 2006; 29: 210 5 24 Meade MO, Granton JT, Matte-Martyn A, et al. A randomized trial of inhaled nitric oxide to prevent ischemia reperfusion injury after lung transplantation. Am J Respir Crit Care Med 2003; 167: 1483 9 25 Nath DS, Walter AR, Johnson AC, et al. Does Perfadex affect outcomes in clinical lung transplantation? J Heart Lung Transplant 2005; 24: 2243 8 26 Oto T, Grifths AP, Rosenfeldt F, et al. Early outcomes comparing Perfadex, Euro-Collins, and Papworth solutions in lung transplantation. Ann Thorac Surg 2006; 82: 1842 8 27 Oto T, Levvey B, Pilcher DV, et al. Evaluation of the oxygenation ratio in the denition of early graft dysfunction post lung transplantation. J Thorac Cardiovasc Surg 2005; 130: 180 6

28 Clark SC. Lung injury after cardiopulmonary bypass. Perfusion 2006; 21: 225 8 29 de Perrot M, Liu M, Waddell TK, Keshavjee S. Ischemia reperfusion-induced lung injury. Am J Respir Crit Care Med 2003; 167: 490 511 30 Fisher AJ, Wardle J, Dark JH, Corris PA. Non-immune acute graft injury after lung transplantation and the risk of subsequent bronchiolitis obliterans syndrome (BOS). J Heart Lung Transplant 2002; 21: 1206 12 31 Abraham D, Taghavi S, Riml P, et al. VEGF-A and -C but not -B mediate increased vascular permeability in preserved lung grafts. Transplantation 2002; 73: 1703 6 32 Schmid RA, Zollinger A, Singer T, et al. Effect of soluble complement receptor type I on reperfusion edema and neutrophil migration after lung allotransplantation in swine. J Thorac Cardiovasc Surg 1998; 116: 90 7 33 Shargall Y, Guenther G, Ahya VN, et al. Report of the ISHLT working group on primary lung graft dysfunction part VI: treatment. J Heart Lung Transplant 2005; 24: 1489 500 34 Vollmar B, Menger MD. Volume replacement and microhemodynamic changes in polytrauma. Langenbecks Arch Surg 2004; 389: 485 91 35 Arif SK, Verheij J, Groeneveld AB, Raijmakers PG. Hypoproteinemia as a marker of acute respiratory distress syndrome in critically ill patients with pulmonary edema. Intensive Care Med 2002; 28: 310 7 36 Groeneveld AB. Radionuclide assessment of pulmonary microvascular permeability. Eur J Nucl Med 1997; 24: 449 61

Downloaded from http://bja.oxfordjournals.org/ by guest on October 28, 2012

514