ORIGINAL ARTICLES

Predictors of Prolonged Postoperative Endotracheal Intubation in Patients Undergoing Thoracotomy for Resection Jacek B. Cywinski, MD,* Meng Xu, MS,† Daniel I. Sessler, MD,‡ David Mason, MD,§ and Colleen Gorman Koch, MD, MS¶

Objective: The aim of this study was to identify predictors sion, higher preoperative serum creatinine level, absence of a of delayed endotracheal extubation defined as the need for thoracic epidural catheter, more extensive surgical resection, postoperative ventilatory support after open thoracotomy and lower preoperative FEV1 were associated with an in- for lung resection. creased risk of delayed extubation after lung resection. Design: An observational cohort investigation. Conclusion: Most predictors of delayed postoperative extu- Setting: A tertiary referral center. bation (ie, red blood cell transfusion, higher preoperative se- Participants: The study population consisted of 2,068 pa- rum creatinine, lower preoperative FEV1, and more extensive tients who had open thoracotomy for , lo- lung resection) are difficult to modify in the perioperative pe- bectomy, or segmental lung resection between January riod and probably represent greater severity of underlying lung disease and more advanced comorbid conditions. However, 1996 and December 2005. thoracic epidural and analgesia is a modifiable fac- Interventions: Not applicable. tor that was associated with reduced odds for postoperative Measurements and Main Results: Preoperative and intra- ventilatory support. Thus, the use of epidural analgesia may operative variables were collected concurrently with the pa- reduce the need for post-thoracotomy . tient’s care. Risk factors were identified using logistic regres- © 2009 Elsevier Inc. All rights reserved. sion with stepwise variable selection procedure on 1,000 bootstrap resamples, and a bagging algorithm was used to KEY WORDS: anesthesia, thoracotomy, postoperative me- summarize the results. Intraoperative red blood cell transfu- chanical ventilation, lung resection, extubation

ELAYED POSTOPERATIVE EXTUBATION prolongs demographic variables, comorbid conditions, laboratory values, and Dthe duration of recovery, increases cost, and may augment operative and outcome variables were collected concurrently with postoperative morbidity.1 Early extubation (defined as extubation patient care by individuals trained in database management in the in the operating room [OR]) decreases resource utilization, im- Department of Cardiothoracic Anesthesiology. Institutional review proves patients’ comfort, reduces morbidity related to mechanical board approval was obtained to perform analyses from the department registry. Perioperative anesthetic care was not dictated by any specific ventilation, and possibly allows for early ambulation and rehabil- protocol. Decisions about patient management were left to the discre- 2,3 itation. After thoracic procedures and especially after lung re- tion of the attending anesthesiologist with the goal to provide optimal section, an additional benefit of limiting positive-pressure mechan- surgical conditions and the ability to extubate at the end of the surgical ical ventilation may include reducing pressure stress on the lung procedure. However, it has been the authors’ practice to initiate epi- tissue suture line, potentially decreasing postoperative air leak.4-6 dural catheters during the surgical procedure to achieve adequate Identifying patients at risk for delayed extubation may allow analgesia at the time of emergence from general anesthesia. clinicians to institute appropriate interventions perioperatively, Descriptive statistics were calculated for all key predictors and demo- which might in turn decrease the duration of postoperative graphic variables. Continuous variables were described with median and ventilatory support and allow for more efficient resource allo- 25th and 75th percentiles, with p values for tests of differences between the cation. The authors’ objective was to identify preoperative 2 groups (extubated in the OR v not extubated in the OR). Categoric variables were described with frequencies and percents, with p values for variables associated with delayed postoperative extubation af- tests of differences between groups. Univariate analysis was performed ter thoracotomy for lung resection. with a chi-square test or Fisher exact test and Wilcoxon rank sum test where appropriate. Categoric outcomes in the two groups were compared METHODS with a chi-square test or Fisher exact test as appropriate. The patient population consisted of 2,068 patients who had open Multivariable logistic regression was used to determine variables that lung resection (pneumonectomy, lobectomy, or segmental lung were associated with postoperative ventilatory support. Risk factors were resection) between January 1996 and December 2005. Preoperative selected by using logistic regression with a stepwise variable selection procedure on 1,000 bootstrap resamples. A bagging algorithm was used to summarize the results. Entry criterion and stay criterion for the stepwise From the Departments of *General Anesthesiology, †Quantitative selection processes were 0.07 and 0.05, respectively. The final logistic Health Sciences, ‡Outcomes Research, §Thoracic and Cardiovascular Sur- regression models for the outcomes were built by using risk factors that gery, and ¶Cardiothoracic Anesthesia, Cleveland Clinic, Cleveland, OH. appeared in 50% of all models from the bootstrap resamples. Address reprint requests to Jacek B. Cywinski, MD, Department of General Anesthesia/E31, Cleveland Clinic, 9500 Euclid Avenue, Cleve- RESULTS land, OH 44195. E-mail: [email protected] © 2009 Elsevier Inc. All rights reserved. Table 1 presents a comparison of patients who were extubated 1053-0770/09/2306-0002$36.00/0 in the OR versus those who arrived to a recovery area intubated, doi:10.1053/j.jvca.2009.03.022 based on categoric preoperative and intraoperative variables. Pa-

766 Journal of Cardiothoracic and Vascular Anesthesia, Vol 23, No 6 (December), 2009: pp 766-769 PROLONGED POSTOPERATIVE ENDOTRACHEAL INTUBATION 767

Table 1. Comparison of Patients Who Were Extubated in the OR Versus Those Who Arrived to a Recovery Area Intubated Based on Categoric Preoperative and Intraoperative Variables

Arrived Extubated Arrived Intubated

Factor Level Total N (%) N (%) p Value*

Male sex 0.58 No 912 788 44.3 124 42.6 Yes 1,156 989 55.7 167 57.4 History of COPD or asthma 0.17 No 1,566 1,355 76.3 211 72.5 Yes 502 422 23.8 80 27.5 History of diabetes 0.11 No 1,820 1,572 88.5 248 85.2 Yes 248 205 11.5 43 14.8 History of stroke 0.023 No 1,978 1,707 96.1 271 93.1 Yes 90 70 3.9 20 6.9 History of end-stage renal disease requiring dialysis 0.99† No 2,061 1,771 99.7 290 99.7 Yes 7 6 0.34 1 0.34 History of smoking 0.62 No 493 427 24.0 66 22.7 Yes 1,575 1,350 76.0 225 77.3 Intraoperative RBC transfusion Ͻ0.001 No 1,857 1,643 92.5 214 73.5 Yes 211 134 7.5 77 26.5 Intraoperative placement of intercostal nerve block 0.46† No 2,064 1,774 99.8 290 99.7 Yes 4 3 0.17 1 0.34 Use of thoracic epidural analgesia Ͻ0.001 No 345 256 14.4 89 30.6 Yes 1,723 1,521 85.6 202 69.4 Use of lumbar epidural analgesia Ͻ0.001 No 1,892 1,637 92.1 255 87.6 Yes 176 140 7.9 36 12.4 Procedure Ͻ0.001 SLR 241 225 12.7 16 5.5 LOL 1,501 1,298 73.0 203 69.8 CP 326 254 14.3 72 24.7 Reintubated Ͻ0.001 No 1,882 1,649 92.8 233 80.1 Yes 186 128 7.2 58 19.9

Abbreviations: LOL, lobectomy of lung; SLR, segmental lung resection; CP, complete pneumonectomy; COPD, chronic obstructive pulmonary disease; RBC, red blood cell. *Chi-square test unless noted. †Fisher exact test. tients who were not extubated in the OR had more intraoperative The final multivariate logistic regression identified the follow- red blood cell (RBC) transfusions, a history of preoperative stroke, ing predictors of postoperative prolonged intubation: intraopera- lower use of thoracic epidural analgesia (TEA), and a higher rate tive RBC transfusion (odds ratio ϭ 2.68 [1.92, 3.74], p Ͻ 0.001), of intraoperative use of lumbar epidural analgesia. A comparison absence of TEA catheter (odds ratio ϭ 2.27 [1.72, 3.03], p Ͻ 0.001), of continuous variables is summarized in Table 2. Patients who elevated preoperative serum creatinine level (odds ratio ϭ 1.42 [1.20, were not extubated in the OR had a lower preoperative forced 1.69], p Ͻ 0.001), and preoperative FEV1 (per 1 unit decrease [odds expiratory volume in 1 second (FEV1), lower forced vital capacity ratio ϭ 1.47 (1.23, 1.72), p Ͻ 0.001]) as well as the extent of the (FVC), lower preoperative hematocrit and albumin, and higher lung resection for segmental lung resection, lobectomy, and com- preoperative creatinine. plete pneumonectomy (more extensive lung resection was predic- The median intubation time in the OR was 3.9 hours (with tive of postoperative ventilation, Table 3 and Fig 1). 25th and 75th percentile of 3.0 and 5.0 hours). The median postoperative intubation time (total intubation time minus the DISCUSSION OR intubation time) was 7.93 hours, with the 25th percentile Early postoperative extubation after lung resection is desir- being 2.75 hours and the 75th percentile 20.67 hours for pa- able because it saves medical resources and may help to avoid tients who were not extubated in the OR. ventilator-related complications.6 An ability to identify periop- 768 CYWINSKI ET AL

Table 2. Comparison of Patients Who Were Extubated in the OR Versus Those Who Arrived to a Recovery Area Intubated Based on Continuous Variables

Arrived Extubated Arrived Intubated

Factor N Median (25th, 75th) N Median (25th, 75th) p Value*

Age (y) 1,777 65.0 (56.5, 72.3) 291 66.6 (57.1, 72.5) 0.24 BSA (m2) 1,777 1.9 (1.7, 2.0) 291 1.8 (1.7, 2.0) 0.21 BMI 1,777 26.3 (23.3, 30.1) 291 26.3 (22.2, 30.5) 0.61

Preoperative Measured FEV1 1,639 2.2 (1.7, 2.8) 267 2.0 (1.6, 2.5) Ͻ0.001 Preoperative Measured FVC 1,639 3.2 (2.5, 4.0) 266 2.9 (2.3, 3.6) Ͻ0.001 Preoperative HCT 1,776 40.5 (37.6, 43.2) 291 39.7 (36.3, 42.8) 0.003 Preoperative serum creatinine (mg/dL) 1,774 0.9 (0.7, 1.0) 291 1.0 (0.8, 1.1) Ͻ0.001 Preoperative albumin (g/dL) 1,738 4.2 (3.9, 4.4) 282 4.1 (3.8, 4.3) Ͻ0.001 Core temperature at the end of the case (°C) 1,720 36.0 (35.5, 36.5) 279 36.0 (35.5, 36.5) 0.81

Abbreviations: BSA, body surface area; BMI, body mass index; HCT, hematocrit. *Wilcoxon sum rank test used. erative modifiable predictors of delayed extubation may help fusion provokes a harmful inflammatory response that may clinicians design interventions to reduce postoperative ventila- contribute to lung injury.12 However, it is more likely that tion time. However, identifying nonmodifiable predictors is higher requirements for intraoperative RBC are surrogates for also helpful because it allows patients to be risk stratified and more complex and technically difficult procedures, which facilitates appropriate allocation of clinical resources. themselves may adversely affect postoperative respiratory out- Patients with poor preoperative lung function had a higher comes.6 Furthermore, the need for RBC transfusion may indi- incidence of postoperative morbidity including prolonged me- cate that patients suffer advanced comorbid conditions, which 7 chanical ventilation. Lower preoperative FEV1 is frequently con- then prompts clinicians to use more liberal thresholds for the sidered a strong predictor of postoperative complications (includ- transfusion as compared with “healthier” patients. ing prolonged mechanical ventilation) after lung resection,8 Adequate pain control after thoracic surgery may improve although it is not uniformly accepted as a predictor of postopera- ventilatory function (better expansion of the chest, deeper res- tive cardiopulmonary complications.6,9-11 In the authors’ patients, piration, and stronger coughing), which contributes to a pa- 13 14 low preoperative FEV1 was associated with greater odds of post- tient’s readiness to be extubated. For example, Bauer et al operative ventilation. However, it seems likely that lower FEV1 is showed that postoperative pain relief was better in patients simply a marker for advanced lung disease, which would be randomized to TEA after lobectomy at rest and on coughing (as difficult to correct with perioperative interventions. Recognizing compared with intravenous morphine) and that there was less the association between low FEV1 and delayed postoperative impairment of FVC and FEV1 postoperatively. It is unknown if extubation may nonetheless help allocate resources (ie, postoper- the same beneficial effect of TEA on FVC and FEV1 can be ative disposition recovery room v intensive care unit) to patients achieved immediately at the end of the surgical procedure, but who are likely to require postoperative mechanical ventilation. it is plausible to assume that reduction in pain and pain-related The present investigation showed that intraoperative RBC impairment of chest mechanics can facilitate earlier extubation. transfusion was strongly predictive of delayed extubation. The Intraoperative activation of an epidural catheter also spares association of intraoperative RBC transfusion with postopera- opioids, which presumably improves postoperative respiratory tive intubation may well be related to the fact that RBC trans- function. Consistent with this theory, the authors found that the use

Table 3. Risk Factors for Delayed Extubation From the Multivariable Regression Model

Factor Odds Ratio (CI) p Value

Intraoperative RBC transfusion 2.68 (1.92, 3.74) Ͻ.001 (yes/no) Preoperative serum creatinine 1.42 (1.20, 1.69) Ͻ0.001 Presence of thoracic epidural 0.44 (0.33, 0.58) Ͻ0.001 Lower preoperative 1.47 (1.23, 1.72) Ͻ0.001

measured FEV1 Type of procedure Ͻ0.001 SLR v CP: 0.33 (0.20, 0.56) Ͻ0.001 LOL v CP: 0.70 (0.52, 0.95) 0.022 SLR v LOL: 0.47 (0.30, 0.75) 0.001 Fig 1. A forest plot showing odds ratio and 95% confidence inter- val of all variables associated with prolonged intubation after open Abbreviations: LOL, lobectomy of lung; SLR, segmental lung resec- thoracotomy for lung resection. LOL, lobectomy of lung; SLR, seg- tion; CP, complete pneumonectomy; RBC, red blood cell; CI, confi- mental lung resection; CP, complete pneumonectomy; RBC, red dence interval. blood cell. PROLONGED POSTOPERATIVE ENDOTRACHEAL INTUBATION 769 of TEA improved the odds of OR extubation in the present morbidity. Clinicians must also consider that factors besides the patients. Other methods of postoperative pain control after thora- extent of lung tissue removed account for the loss in respiratory cotomy have been reported in the literature including paravertebral function observed early after lung resection including impairment and intercostal blocks and interpleural and subarachnoid adminis- in chest wall compliance, accumulated bronchial secretions, bron- tration of drugs. Among these alternatives, paravertebral blocks chial hyperreactivity, microatelectasis, increased lung water content, appear to be the most effective alternative when epidural blocks diaphragmatic dysfunction, and reduced surfactant activity.17,18 cannot be performed.15,16 Thus, it seems likely that a paravertebral The main limitation of the present analysis is its retrospec- block would provide benefits similar to TEA. tive nature, which could not account for all variables, poten- Preoperative renal dysfunction is a well-recognized risk for in- tially affecting decisions of when to extubate the patient. Fur- 7 creased morbidity after cardiac and thoracic surgery. Renal impair- thermore, patients in the present study represent a population ment is considered by many as a marker of cardiovascular dysfunc- treated over a period of 9 years; hence, changes in the practice tion and by itself can make perioperative fluid management more could play a role in the decision regarding timing of extubation. difficult, potentially affecting lung function after resection.7 As And, finally, the authors were unable to analyze the association might be expected, elevated preoperative creatinine was associated between predicted postoperative FEV and delayed extubation, with delayed endotracheal extubation in the present patients. 1 which might have been, as some suggest, a stronger predictor The authors found that the volume of the surgical lung resection of postresection lung function than preoperative FEV alone. was a good predictor of delayed postoperative extubation (the 1 more extensive the resection, the greater the chance for postoper- In summary, the authors found most predictors of delayed ative mechanical ventilation). However, it would be an oversim- endotracheal extubation to be difficult to correct in the perioper- plification to assume that a larger volume of resected lung tissue ative period. Lower FEV1, elevated preoperative creatinine, and alone predisposes to extended mechanical ventilation because in intraoperative requirements for RBC transfusion may simply rep- many cases the resected, diseased portion of the lung does not resent a more advanced underlying lung disease, advanced comor- participate in respiratory function. That fact can explain why in the bid conditions, or technically more difficult surgical procedures. present analysis preoperative FEV1 was strongly associated with Recognizing the predictive importance of these factors may none- the chance for postoperative ventilation without adjustment for the theless help identify patients at risk for extended postoperative loss of lung function after resection. mechanical ventilation and help to allocate postoperative re- Of course, extensive lung resection is also likely to be a surrogate sources. In contrast, the use of TEA is a straightforward interven- for a more advanced and/or extensive disease process, which tion that may reduce the odds of prolonged postoperative ventila- presumably contributes independently to postoperative respiratory tion.

REFERENCES 1. Desiderio D, Downey R: Critical issues in early extubation and 11. Kearney DJ, Lee TH, Reilly JJ, et al: Assessment of operative hospital discharge in thoracic oncology surgery. J Cardiothorac Vasc risk in patients undergoing lung resection. Importance of predicted Anesth 12:3-6, 1998 pulmonary function. Chest 105:753-759, 1994 2. Dos Santos CC, Slutsky AS: Invited review: Mechanisms of ventilator- 12. Fransen E, Maessen J, Dentener M, et al: Impact of blood induced lung injury: A perspective. J Appl Physiol 89:1645-1655, 2000 transfusions on inflammatory mediator release in patients undergoing 3. Dos Santos CC, Slutsky AS: Mechanotransduction, ventilator- . Chest 116:1233-1239, 1999 induced lung injury and multiple organ dysfunction syndrome. Inten- 13. Ballantyne JC, Carr DB, deFerranti S, et al: The comparative sive Care Med 26:638-642, 2000 effects of postoperative therapies on pulmonary outcome: 4. Wright CD, Wain JC, Mathisen DJ, et al: Postpneumonectomy Cumulative meta-analyses of randomized, controlled trials. Anesth bronchopleural fistula after sutured closure: Incidence, risk fac- tors and management. J Thorac Cardiovasc Surg 112:1367-1371, 1996 Analg 86:598-612, 1998 5. Auriant I, Jallot A, Hervé P, et al: Noninvasive ventilation re- 14. Bauer C, Hentz JG, Ducrocq X, et al: Lung function after duces mortality in acute following lung resection. lobectomy: A randomized, double-blinded trial comparing thoracic Am J Respir Crit Care Med 164:1231-1235, 2001 epidural ropivacaine/sufentanil and intravenous morphine for patient- 6. Stéphan F, Boucheseiche S, Hollande J, et al: Pulmonary com- controlled analgesia. Anesth Analg 105:238-244, 2007 plications following lung resection: A comprehensive analysis of inci- 15. Joshi GP, Bonnet F, Shah R, et al: A systematic review of dence and possible risk factors. Chest 118:1263-1270, 2000 randomized trials evaluating regional techniques for postthoracotomy 7. Wright CD, Gaissert HA, Grab JD, et al: Predictors of prolonged analgesia. Anesth Analg 107:1026-1040, 2008 length of stay after lobectomy for : A Society of Thoracic 16. Fibla JJ, Molins L, Mier JM, et al: Comparative analysis of Surgeons General Thoracic Surgery Database risk-adjustment model. analgesic quality in the postoperative of thoracotomy: Paravertebral Ann Thorac Surg 85:1857-1865, 2008 block with bupivacaine 0.5% vs ropivacaine 0.2%. Eur J Cardiothorac 8. Licker MJ, Widikker I, Robert J, et al: Operative mortality and Surg 33:430-434, 2008 respiratory complications after lung resection for cancer: Impact of chronic obstructive pulmonary disease and time trends. Ann Thorac 17. Sedrakyan A, Zhang H, Treasure T, et al: Recursive partition- Surg 81:1830-1837, 2006 ing-based preoperative risk stratification for atrial fibrillation after 9. Linden PA, Bueno R, Colson YL, et al: Lung resection in patients coronary artery bypass surgery. Am J 151:720-724, 2006 Ͻ with preoperative FEV1 35% predicted. Chest 127:1984-1990, 2005 18. Varela G, Jiménez MF, Novoa N, et al: Discordance between 10. Varela G, Brunelli A, Rocco G, et al: Measured FEV1 in the first predicted postoperative forced expiratory volumes in one second postoperative day, and not ppoFEV1, is the best predictor of cardio-respiratory (ppoFEV1) calculated before and after resection of bronchogenic car- morbidity after lung resection. Eur J Cardiothorac Surg 31:518-521, 2007 cinoma. Interact Cardiovasc Thorac Surg 2:138-142, 2003