Secondary Logo

Journal Logo

Outcome

Risk prediction model for respiratory complications after lung resection

An observational multicentre study

Yepes-Temiño, Maria J.; Monedero, Pablo; Pérez-Valdivieso, José Ramón Grupo Español de Anestesia Toracica

Author Information
European Journal of Anaesthesiology: May 2016 - Volume 33 - Issue 5 - p 326-333
doi: 10.1097/EJA.0000000000000354
  • Free

Abstract

Introduction

Postoperative pulmonary complications (PPCs) are one of the most important causes of morbidity and mortality after thoracic surgery.1,2 Despite improvements in patient selection, anaesthesia, ventilatory management, surgical techniques and intensive care management, the overall incidence of PPCs after thoracic surgery varies from 10 to 37%, depending on the design of the study.3–5 A validated model for risk stratification of PPCs would be useful at preoperative assessment of patients undergoing thoracic surgery for guiding clinical decision-making, intervening on modifiable factors and informing the patient about the perioperative risks.6 The accurate identification of high-risk patients could also guide ICUs in distributing resources and facilitate the selection of patients for clinical trials.7

Several logistic models and scores have been developed and tested to improve stratification of mortality risk in thoracic surgery,8,9 but few studies have dealt only with pulmonary complication risks. Amar et al.10 published a Clinical Prediction Rule for Pulmonary Complications (CPRPCs) after thoracic surgery for primary lung cancer, exclusively based on previous chemotherapy (point score 2) and lower predicted postoperative (PPO) diffusing capacity of the lung for carbon monoxide (DLCO%) (point score 1 for each 5% decrement of DLCO <100%). They defined three overall risk categories for PPCs: low less than 10 points (estimated PPCs 9%); intermediate 11 to 13 points (PPCs 14%); and high more than 14 points (PPCs 26%).

None of the predictive models available has been studied in settings outside the development context to test transportability, which is a problem in risk prediction. Unless prediction scores are clearly transportable and performing as expected in new samples of patients, generalised use cannot be warranted. Therefore, reliable PPC prediction continues to represent an important knowledge gap with practical implications. The primary aim of our study was to provide external validation of the CPRPC after lung resection for primary tumours. The hypothesis was that CPRPC performs well for prediction of PPCs. If there was poor discrimination, we planned, as a second objective, to derive a new predictive index for PPCs.

Patients and methods

This multicentre, retrospective observational study was performed in a cohort of all consecutive adult patients who underwent pulmonary resection in 13 Spanish hospitals from 1 January to 30 June 2011. Data from 559 consecutive patients were collected and analysed. Data were obtained from the clinical history and computer records from admission to the thoracic surgery department until 30 days after surgery (Fig. 1).

Inclusion criteria were adult patients (over 17 years old) who had undergone any kind of lung resection (oncological or lung volume reduction surgery), thoracotomy or videothoracoscopy. We collected the same demographic, clinical and hospital outcome data, including the presence of PPCs, as defined in the prior research to calculate the CPRPC.10 The definitions of PPCs were respiratory failure requiring ICU admission and/or tracheal intubation, pneumonia (fever and new pulmonary infiltrate treated with intravenous antibiotics), atelectasis requiring bronchoscopy (need determined by the surgical or ICU team), pulmonary embolism (diagnosed by imaging technique and treated) and need for supplemental oxygen on hospital discharge. The risk score and probability of PPCs were calculated for each patient. We used the following equation to predict postoperative lung function:

Postoperative function = preoperative measured lung function × [1 – (number of segments resected/total number of segments)].11

Statistical analysis

The outcome variable was the presence of at least one PPC, as defined by CPRPC. Normally distributed continuous data are presented as mean ± SD and compared using the Student's t test. Nonnormally distributed variables are shown as median (10th to 90th percentiles). Categorical variables are expressed as proportions and compared using Fisher's exact test. We compared our cohort with that of CPRPC using the χ2 test for categorical variables. The performance of the CPRPC was determined by examining its ability to discriminate and calibrate. Discriminatory power was tested using the area under the receiver-operating characteristic (AUROC) curve. Calibration was assessed using the Hosmer-Lemeshow goodness-of-fit statistic. Patients with missing data were excluded from the analyses for CPRPC validation.

After examining the performance of the CPRPC in the subpopulation of patients with primary tumours equivalent to the study of Amar et al.,10 we developed a new predictive index for PPCs in our cohort. We used the potential PPC predictor variables according to earlier studies. A logistic regression model was constructed based on the presence of a PPC (as defined by CPRPC) as the dependent variable. Bivariate analysis was first performed to identify the preoperative variables that were associated (P < 0.05) with PPCs. Multivariate linear regression modelling was performed to assess the adjusted associations of the variables with the occurrence of postoperative complications. All clinically appropriate, based on previous literature, or statistically significant variables were considered for inclusion in the model.5,10 The nature of the relationships between predictor variables and complications was also compared among hospitals. Modelling was performed by backward stepwise selection, for which P less than 0.05 was the criterion for variable retention. The significance of the interactions of explanatory variables (as guided by the literature and clinical judgement) was assessed through a product term included in the multivariate models. Independent prediction factors were assigned a value, based on their β-coefficient in the logistic regression analysis. Each patient was assigned a total number of points and a predictor value using the model. Sensitivity and specificity were calculated for each possible value. AUROC was computed as a descriptive tool for measuring discrimination by the model. The Hosmer-Lemeshow goodness-of-fit statistic was calculated to examine the fit of the model. To perform an internal validation of the proposed score, we conducted bootstrapping with 1000 replications using bias-corrected confidence intervals (CIs).12 Statistical analyses were performed using SPSS, version 15.0 (SPSS Inc., Chicago, Illinois, USA) and STATA (version 12; Stata Corp, College Station, Texas, USA).

Results

Demographic and operative characteristics of the 559 patients of our cohort who underwent pulmonary resection are shown in Table 1. Two-thirds of patients had primary lung cancer (n = 359), 25% had lung metastatic disease and 10% had nononcological disease. The median age was 64 years and 4% of the patients were older than 80 years. Only 41 (7.3%) patients had preoperative chemotherapy.

Table 1
Table 1:
Demographic and clinical characteristics of patients studied

The overall rate of PPCs in the whole cohort was 11.6% (n = 65). Females and younger patients had a significantly lower incidence of PPCs. Pneumonectomy was associated with the highest incidence of PPCs (19%) and lobectomy was the most common surgery, performed in 302 (54%) patients. Former (stopped two or more weeks before surgery) and current smokers, higher haemoglobin concentration and chronic obstructive pulmonary disease were associated with the presence of PPCs. With regard to respiratory function tests, the mean percentage forced expiratory volume in 1 s (FEV1) and the PPO FEV1 (FEV1%) were lower in patients who suffered PPCs (Table 1). The incidence of PPCs was higher in patients with primary lung cancer compared with pulmonary metastasis or nononcological disease.

Patients with PPCs had worse outcomes and increased mortality rates (Table 2). The median postoperative length of hospital stay and ICU stay were longer in patients with PPCs. There were six in-hospital deaths (1.1% of the population and 9.2% of those who developed PPCs), all of whom had developed PPCs. The final samples included in the statistical analysis are described in Fig. 1.

Table 2
Table 2:
Outcomes relating to postoperative pulmonary complications
Fig. 1
Fig. 1:
Study flow chart. The population for the Clinical Prediction Rule for Pulmonary Complication (CPRPC) validation included all patients with primary lung cancer and with full information about diffusing capacity of the lung for carbon monoxide (DLCO%). All patients were included for the development of the new score. PPC, postoperative pulmonary complication.

Of the 359 patients who had primary lung cancer, we excluded 135 patients for whom PPO DLCO% had not been recorded, obtaining a cohort of 224 patients with 36 suffering PPCs. The incidence of one or more PPCs in this subset of patients was similar to the Amar et al.'s10 population (16.1 vs. 12.6%, respectively, two-tailed Fisher's exact test P = 0.19). Thirty-one patients (13.8%) developed respiratory failure, 14 (6.2%) had pneumonia, 14 (6.2%) suffered atelectasis requiring bronchoscopy, three (1.3%) had pulmonary embolism and seven (3.1%) needed supplemental oxygen on hospital discharge. The univariate comparisons of patients with or without PPCs are shown in Table 3. Preoperative chemotherapy was given to 19 (8.5%) patients: 18 of these did not have PPCs. According to the CPRPC, our cohort included 19 patients in the high-risk category (11% suffered PPCs), 49 patients in the intermediate-risk category (20% presenting PPCs) and 156 low-risk patients (15.4% presented PPCs). We did not find any significant association between PPCs and PPO DLCO% or preoperative chemotherapy. The AUROC using the CPRPC showed poor discrimination when applied to the primary lung cancer population (0.47; 95% CI 0.37 to 0.57) (Fig. 2). The Hosmer-Lemeshow goodness-of-fit statistic suggested good calibration (P = 0.7), but calibration is only useful with regard to discrimination; predictive models that perform no better than chance will perform consistently across all risk strata and may show excellent calibration.

Table 3
Table 3:
Univariate comparisons in subsample equivalent to population of Amar et al. 10
Fig. 2
Fig. 2:
. Receiver-operating characteristic (ROC) curve plot for the Clinical Prediction Rule for Pulmonary Complication risk score in the subpopulation of patients with primary tumour in our cohort. Area under the ROC curve 0.47 (95% confidence interval 0.37 to 0.57).

As a result of the poor fit of the CPRPC model to our data, we decided to generate a new score that would discriminate between the risks of our different subpopulations by fitting a stepwise multivariate logistic regression. The variables introduced in the multivariate model were sex, American Society of Anesthesiologists’ physical status, age, BMI, haemoglobin concentration, smoking status (no smoker, former smoker, current smoker), preoperative respiratory symptoms, preoperative SpO2, preoperative FEV1%, PPO FEV1%, PPO DLCO%, preoperative chemotherapy, surgical procedure (wedge, lobectomy, pneumonectomy), volume of intraoperative fluids and type of tumour (primary tumour, metastasis, no tumour). We found that being a current smoker or a former smoker, the age and the PPO FEV1% were statistically significant factors for predicting PPCs in our cohort (Table 4). Moreover, a simplified predictive risk score was calculated by multiplying each logistic coefficient of regression (β) by 30 and rounding off its value as follows:

Table 4
Table 4:
Odds ratios and 95% confidence intervals (CI) for postoperative pulmonary complications

Age (years) – PPO FEV1 (%) + (50 if current smoker or 35 if ex-smoker)

The median predictive risk score in our sample was 27.6 (range −70.9 to 88.4). We selected a score of 30 as the optimal cut-off value, based on the highest specificity and sensitivity (sensitivity 80% and specificity 61%). According to that, 63% of our cohort were successfully classified as high risk with a likelihood ratio of 2.02. The performance of the new score proved to be satisfactory, with an AUROC of 0.74 (95% CI 0.70 to 0.78), suggesting good discriminatory ability and better than CPRPC (Fig. 3). The Hosmer-Lemeshow goodness-of-fit test, using deciles to group the data for testing, suggested good calibration (P = 0.26). The average value of the difference between expected complications using the predictive risk score and observed data was 1.85 ± 1.86. The internal validation of the new score, relying on 1000 bootstrap resamples, showed a bootstrap-corrected predictive accuracy of 0.74 (95% CI 0.69 to 0.79).

Fig. 3
Fig. 3:
Receiver-operating characteristic (ROC) curve plot for the new risk score in our whole cohort. Area under the ROC curve 0.74 (95% confidence interval 0.70 to 0.78).

Discussion

Our study is the first attempt at external validation of the CPRPC proposed by Amar et al.10 We found that, despite its simplicity, it poorly performed when applied to our cohort of patients, with a low discriminatory power in primary lung cancer surgical patients. We did not find any significant association between either PPO DLCO% or preoperative chemotherapy and the presence of PPCs. However, age, smoking status and PPO FEV1% were strong independent factors for PPCs prediction.

A reason for the poor performance of CPRPC could be related to differences between the studied populations (even though the incidence of PPCs was not significantly different) or related to perioperative management. The CPRPC was obtained from a retrospective review of a database of 956 patients who suffered lung resection at a single institution between 1992 and 2003. During those 11 years, there were medical progresses that precluded a homogeneous cohort.13 PPCs occurred in 12.7% of patients. The two independent risk factors for PPCs in the CPRPC were preoperative chemotherapy and a lower PPO DLCO%.

According to our results, PPO DLCO% was not a good predictor of PPCs, whereas PPO FEV1% was an effective independent predictor. FEV1 is considered a principal prognosis factor in the guidelines of the American College of Chest Physicians, British Thoracic Society and European Respiratory Society/European Society of Thoracic Surgeons. Postoperative FEV1% has been previously described as a risk factor for postoperative morbidity with a lengthy ICU stay in patients who underwent lung resection surgery.14 The European Society Objective Score, involving age and postoperative FEV1%, was developed to identify preoperative risk factors associated with mortality after lung resection surgery.8 There is poor correlation between FEV1 and DLCO because they measure different aspects of lung function.15–17 Prethoracotomy respiratory assessment includes parenchymal function (DLCO), lung mechanics (FEV1) and cardiopulmonary reserve (VO2max), but estimation of suitability for lung resection is a different issue from PPC prediction. Our cohort included only patients who were considered fit for surgery after a preoperative assessment and this may have resulted in fewer patients with poor lung function undergoing surgery.18 Furthermore, DLCO was performed only in patients with limited exercise tolerance or reduced lung volumes. Weaknesses of FEV1 and DLCO must be considered when used alone to stratify the risk of pulmonary surgery.13,19–22

The impact of neo-adjuvant chemotherapy on morbidity and mortality after lung resection is also disputable. We did not find any relationship between neo-adjuvant chemotherapy and the incidence of PPCs. The incidence of preoperative chemotherapy in our cohort was lower than in the original publication (8 vs. 20%; two-tailed Fisher's exact test P < 0.0001), which could reflect a difference in current clinical practice. Our cohort is more contemporary and follows the present recommendations and evidence for preoperative chemotherapy. In the largest trial reported of preoperative chemotherapy in lung cancer, 519 patients were randomly assigned to receive either surgery alone or three cycles of platinum-based chemotherapy followed by surgery.23 Most patients (61%) had clinical stage I disease, 31% stage II and 7% stage III disease. No advantage in survival was seen from neo-adjuvant chemotherapy in any of the stages, and postoperative complications were similar between groups. In addition, our whole cohort included patients with metastatic disease who received chemotherapy regimens that differ from those for primary lung cancer. Finally, patients with primary lung cancer developed more complications than those who underwent metastasectomy. The uncertain prognosis of patients with metastatic lung disease may foster parenchyma-sparing procedures, in contrast with primary lung cancer that usually requires extended resections.24

Owing to the poor performance of the CPRPC, we developed a new model to classify the patients in our cohort. Based on the data collected, a simple PPC risk stratification scale for patients scheduled for lung resection was developed, including tests obtained before surgery. Age, smoking status and PPO FEV1% contributed independently to the risk calculation for clinically important PPCs. Our new risk stratification system is suitable for all patients who undergo lung resection, including those scheduled for resection of metastases and nononcological patients. These three risk factors are easily available before surgery, and the system is based solely on preoperative variables, making it convenient.6 Advanced age is associated with poorer functional status and a higher incidence of coexisting diseases.25 Several health conditions in the analysis, such as hypertension, impaired renal function or liver disease, were unrelated to PPCs.26,27 Smoking is associated with an increased risk of PPCs. As in studies performed by other groups, we divided our population into three categories: current smokers, former smokers and nonsmokers.28,29 We found an increased risk of PPCs in patients who smoked at the time of surgery, as previously described.30 Although the duration of smoking abstinence necessary for a reduction in PPCs has not been firmly established, preoperative smoking cessation is a recommended strategy.

The current study has important strengths. It comprises a large population, multicentre based, with accurate healthcare data collected in the actual practice circumstances of pulmonary resection surgery. Our cohort included more contemporary patients than the study of Amar et al.10 so the recorded events and procedures are more up to date regarding advances in treatment guidelines, surgical techniques and present intensive care management. The outcome of interest (PPCs) was clearly defined and clinically relevant with the use of hard and stringent definitions identical to those of Amar et al.10 The logistic regression analyses adhered to sample size recommendations for 10 or more outcome events per predictor variable. Our score can be applied to all patients scheduled for lung resection to identify high-risk patients of PPCs.

The main clinical application of the current study is that clinicians should not use the CPRPC to estimate the risk of PPCs after lung resection surgery. We offer a simple new score that recognises older patients, current or former smokers, with low estimated postoperative FEV1% as a higher-risk group of patients who need specific information and special perioperative care. These patients could benefit from preoperative optimisation or the consideration of nonsurgical alternatives. Smoking represents a potentially avoidable and modifiable risk factor for PPCs.31 Potential key interventions likely to reduce the risk of PPCs would be protective ventilation, restrictive intravascular volume optimisation, pain management with regional instead of systemic analgesia, aggressive postoperative physiotherapy and early mobilisation.32 Physiotherapy, incentive spirometry with inspiratory muscle training and prophylactic minitracheostomy have all been shown to reduce the frequency of PPCs. Early use of noninvasive ventilation may reduce the need for endotracheal mechanical ventilation and mortality.30

There are a few limitations to be considered when interpreting our results. A large variety of researchers collecting data might increase the chance of errors in data acquisition. Although we included many factors related to PPCs such as protective ventilation, intraoperative fluids or blood transfusion, absence of data on other potential confounding factors, such as postoperative physiotherapy or early mobilisation after surgery, may have biased our results. A larger number of patients would have allowed an increased number of predictor variables. Despite our internal validation with a multivariate 1000 bootstrap-corrected predictive accuracy of 74%, a further step in the proof of our score will require external validation.

In summary, this multicentre retrospective study shows that the CPRPC is not a good predictor of PPCs in our population. In contrast, three significant independent preoperative risk factors for PPCs in lung resection were identified in our cohort: older age; current or former smoking status; and reduced PPO FEV1%. A score of 30 or more predicts a high risk of PPCs. We have developed an easy and inexpensive scoring system to predict PPCs in lung resection candidates that will require further external validation. The new score may detect high-risk patients who may be candidates for recruitment into clinical trials of novel pulmonary protective therapies or benefit from more intensive care.

Acknowledgements relating to this article

Assistance with the study: we are grateful to Maria Canales-Gortazar for language editing.

Financial support and sponsorship: none.

Conflicts of interest: none.

Presentation: none.

Appendix 1

Grupo Español de Anestesia Toracica

María José Yepes, Pablo Monedero, Maira Bes-Rastrollo (Clínica Universidad de Navarra), José Ramón Pérez-Valdivieso (Complejo Hospitalario de Navarra), Maira Bes-Rastrollo (Universidad de Navarra) Manuel Granell-Gil, Ricardo Guijarro-Jorge, Santiago Figueroa Almánzar, Conrado Mínguez Marín, Antonio Arnau Obrer (Consorcio Hospital General Universitario de Valencia), María-José Jiménez, Clara Hernández Clara, Marc Giménez-Milà (Hospital Clínic de Barcelona), Patricia Ruiz-Granado, Elisa Álvarez Fuente, Eduardo Tamayo (Hospital Clínico Universitario de Valladolid), Rafael Anaya-Camacho, María Carmen Unzueta Merino (Hospital de la Santa Creu i Sant Pau de Barcelona), Patricia Piñeiro, María Palencia, Patricia Duque, Estrella Terradillos, Beatriz Bunger, Ana Lajara-Montell (Hospital General Universitario Gregorio Marañón de Madrid), Rosa Villalonga-Vadell, Albert Pi-López, Miguel Ángel Delgado-Carrasco, Juan Moya-Amorós, Anna Ureña-Lluveras, Gabriela Rosado-Rodríguez (Hospital Universitari de Bellvitge), Manuel Barbera, Yolanda Fernández-Fernández, Rosario Vicente (Hospital Universitari i Politècnic La Fe de Valencia), Belén Pérez-Cámara, Alazne Enparantza-Aiestaran, Carma Queralt-Pascual, María Eizaguirre-Cotado, Edurne Lodoso-Ochoa and Orieta Silvera-Soto (Hospital Universitario Donostia de San Sebastián), Blanca Hermosa-Contreras, María García-Sáez, Isabel A. Becerra-Cayetano, Santiago García-Barajas, Isabel Gragera-Collado (Hospital Universitario Infanta Cristina de Badajoz), Marcos Martínez-Borja, Diego Parise-Roux, Elena Elias-Martín, Pilar Arribas-Pérez, Noemi Samaranch-Palero and Cristina Carrasco-Seral (Hospital Universitario Ramón y Cajal de Madrid), Antonio Ramón Fernández-López, José M. López-Millán (Hospital Universitario Virgen Macarena de Sevilla), Silvia Bermejo-Martínez, Alejandro Pérez-Ramos, Enrique Vela. Fernando Escolano-Villén (Parc de Salut MAR de Barcelona).

References

1. Johnson RG, Arozullah AM, Neumayer L, et al. Multivariable predictors of postoperative respiratory failure after general and vascular surgery: results from the patient safety in surgery study. J Am Coll Surg 2007; 204:1188–1198.
2. Smetana GW, Lawrence VA, Cornell JE. American College of Physicians. Preoperative pulmonary risk stratification for noncardiothoracic surgery: systematic review for the American College of Physicians. Ann Intern Med 2006; 144:581–595.
3. Canet J, Gallart L, Gomar C, et al. Prediction of postoperative pulmonary complications in a population-based surgical cohort. Anesthesiology 2010; 113:1338–1350.
4. Allen MS, Darling GE, Pechet TT, et al. Morbidity and mortality of major pulmonary resections in patients with early-stage lung cancer: initial results of the randomized, prospective ACOSOG Z0030 trial. Ann Thorac Surg 2006; 81:1013–1020.
5. Canet J, Gallart L. Predicting postoperative pulmonary complications in the general population. Curr Opin Anaesthesiol 2013; 26:107–115.
6. Barnett S, Moonesinghe SR. Clinical risk scores to guide perioperative management. Postgrad Med J 2011; 87:535–541.
7. Pace NL, Eberhart LH, Kranke PR. Quantifying prognosis with risk predictions. Eur J Anaesthesiol 2012; 29:7–16.
8. Berrisford R, Brunelli A, Rocco G, et al. The European thoracic surgery database project: modelling the risk of in-hospital death following lung resection. Eur J Cardiothorac Surg 2005; 28:306–311.
9. Falcoz PE, Conti M, Brouchet L, et al. The thoracic surgery scoring system (thoracoscore): risk model for in-hospital death in 15 183 patients requiring thoracic surgery. J Thorac Cardiovasc Surg 2007; 133:325–332.
10. Amar D, Munoz D, Shi W, et al. A clinical prediction rule for pulmonary complications after thoracic surgery for primary lung cancer. Anesth Analg 2010; 110:1343–1348.
11. Nakahara K, Ohno K, Hashimoto J, et al. Prediction of postoperative respiratory failure in patients undergoing lung resection for lung cancer. Ann Thorac Surg 1988; 46:549–552.
12. Harrell FE Jr, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med 1996; 15:361–387.
13. Lim E, Baldwin D, Beckles M, et al. Guidelines on the radical management of patients with lung cancer. Thorax 2010; (65 Suppl 3):iii1–iii27.
14. Yang M, Ahn HJ, Kim JA, Yu JM. Risk score for postoperative complications in thoracic surgery. Korean J Anesthesiol 2012; 63:527–532.
15. Ferguson MK, Little L, Rizzo L, et al. Diffusing capacity predicts morbidity and mortality after pulmonary resection. J Thorac Cardiovasc Surg 1988; 96:894–900.
16. Mazzone P. Preoperative evaluation of the lung resection candidate. Cleve Clin J Med 2012; 79 (Electronic Suppl 1):eS17–eS22.
17. Brunelli A, Refai MA, Salati M, et al. Carbon monoxide lung diffusion capacity improves risk stratification in patients without airflow limitation: evidence for systematic measurement before lung resection. Eur J Cardiothorac Surg 2006; 29:567–570.
18. Colice GL, Shafazand S, Griffin JP, et al. Physiologic evaluation of the patient with lung cancer being considered for resectional surgery: ACCP evidenced-based clinical practice guidelines (2nd edition). Chest 2007; (132 Suppl 3):161S–177S.
19. Brunelli A, Charloux A, Bolliger CT, et al. ERS/ESTS clinical guidelines on fitness for radical therapy in lung cancer patients (surgery and chemo-radiotherapy). Eur Respir J 2009; 34:17–41.
20. Berry MF, Hanna J, Tong BC, et al. Risk factors for morbidity after lobectomy for lung cancer in elderly patients. Ann Thorac Surg 2009; 88:1093–1099.
21. Ferguson MK, Lehman AG, Bolliger CT, Brunelli A. The role of diffusing capacity and exercise tests. Thorac Surg Clin 2008; 18:9–17.
22. Salati M, Brunelli A. Preoperative assessment of patients for lung cancer surgery. Curr Opin Pulm Med 2012; 18:289–294.
23. Gilligan D, Nicolson M, Smith I, et al. Preoperative chemotherapy in patients with resectable nonsmall cell lung cancer: results of the MRC LU22/NVALT 2/EORTC 08012 multicentre randomised trial and update of systematic review. Lancet 2007; 369:1929–1937.
24. Welter S, Cheufou D, Ketscher C, et al. Risk factors for impaired lung function after pulmonary metastasectomy: a prospective observational study of 117 cases. Eur J Cardiothorac Surg 2012; 42:e22–e27.
25. Banz VM, Jakob SM, Inderbitzin D. Improving outcome after major surgery: pathophysiological considerations. Anesth Analg 2011; 112:1147–1155.
26. Arozullah AM, Khuri SF, Henderson WG, Daley J. Participants in the National Veterans Affairs Surgical Quality Improvement Program. Development and validation of a multifactorial risk index for predicting postoperative pneumonia after major noncardiac surgery. Ann Intern Med 2001; 135:847–857.
27. Stephan F, Boucheseiche S, Hollande J, et al. Pulmonary complications following lung resection: a comprehensive analysis of incidence and possible risk factors. Chest 2000; 118:1263–1270.
28. Barrera R, Shi W, Amar D, et al. Smoking and timing of cessation: impact on pulmonary complications after thoracotomy. Chest 2005; 127:1977–1983.
29. Raupach T, Quintel M, Hinterthaner M. Preoperative smoking cessation in patients with lung cancer. Pneumologie 2010; 64:694–700.
30. Agostini P, Cieslik H, Rathinam S, et al. Postoperative pulmonary complications following thoracic surgery: are there any modifiable risk factors? Thorax 2010; 65:815–818.
31. Thomsen T, Villebro N, Moller AM. Interventions for preoperative smoking cessation. Cochrane Database Syst Rev 2014; 3:CD002294.
32. Guldner A, Pelosi P, de Abreu MG. Nonventilatory strategies to prevent postoperative pulmonary complications. Curr Opin Anaesthesiol 2013; 26:141–151.
© 2016 European Society of Anaesthesiology