A model to predict the decline of the forced expiratory volume in one second and the carbon monoxide lung diffusion capacity early after maj
http://www.100md.com
《血管的通路杂志》
Unit of Thoracic Surgery, ‘Umberto I°’ Regional Hospital, Ancona, Italy
Abstract
The objective of the study was to develop regression models for the prediction of the decline of the forced expiratory volume in one second (FEV1) and the carbon monoxide lung diffusion capacity (DLCO) early after major lung resection. One hundred and ninety patients submitted to pulmonary lobectomy or pneumonectomy for lung cancer performed preoperative and early postoperative (mean 10.9 after operation) pulmonary function tests. One hundred and fifty of these patients also underwent DLCO measurements by the single breath method. The decline of FEV1 and DLCO were expressed as percentage losses from preoperative values. Stepwise multiple regression analyses were performed to develop two models estimating the percent reduction of FEV1 and DLCO from preoperative values. The multivariate procedures were then validated by bootstrap analyses. The following regression equations were derived: estimated percent reduction in FEV1=21.34 – (0.47xage) + (0.49xpercentage of functioning parenchyma removed during operation) + (17.91xCOPD-index); estimated percent reduction in DLCO= 35.99 – (0.31xage) – (36.47xFEV1/FVC ratio) + (0.33xDLCO) + (0.54xpercentage of functioning parenchyma removed during operation). The comparison between observed and estimated losses of FEV1 and DLCO (by using these regression equations) was not significantly different. We think the regression models generated in this study may be reliably used for risk stratification purposes.
Key Words: Lung cancer; Lung resection; Pulmonary function test; Lung diffusion capacity; Postoperative period; Regression analysis
1. Introduction
Major lung resections determine a variable reduction of pulmonary function after operation. Particularly in the early postoperative period, this reduction may be of such extent to cause major morbidity, and hinder the postoperative recovery.
Therefore, the objective of the present study was to develop regression models for the prediction of the decline of two commonly used respiratory parameters, such as the forced expiratory volume in one second (FEV1) and the carbon monoxide lung diffusion capacity (DLCO).
2. Patients and methods
Two hundred and thirty-six patients submitted to lobectomy or pneumonectomy for non small cell lung cancer (NSCLC) from October 2001 through April 2004 were enrolled in our study after giving informed consent. Eight patients undergoing lung resection associated with chest wall resection were excluded. The other 38 patients were excluded from the study because they did not perform the postoperative pulmonary function tests (PFTs) (causes: 10 postoperative deaths, 14 patients with prolonged air leaks discharged with chest tubes and Heimlich valves, 3 patients with empyema and 3 with pneumonia transferred to other local hospitals for further assistance, 2 patients with bronchopleural fistula, 2 patients with postoperative myocardial infarction, 2 patients with postoperative strokes, 1 patient with pulmonary embolism and 1 refusal to perform the PFT).
Thus, the remaining 190 patients (151 males, 39 females; 161 pulmonary lobectomy, L; and 29 pneumonectomy, P) formed the database of the study.
Our criteria for inoperability were a predicted postoperative FEV1 (ppoFEV1) less than 30% of predicted in association with a height climbed at the stair climbing test lower than 12 m, and/or hemodynamic instability [1].
All patients performed lung resection through a muscle-sparing lateral thoracotomy. The following procedures were performed: 40 left upper lobectomies, 39 right upper lobectomies, 37 left lower lobectomies, 24 right lower lobectomies, 4 middle lobectomies, 10 lower bilobectomies, 7 upper bilobectomies, 16 left pneumonectomies, 13 right pneumonectomies.
Postoperative respiratory treatment consisted of daily chest physiotherapy administered by a specialized physiotherapist, early mobilization, incentive spirometer for all patients. Bronchodilators were administered only when clinically indicated. Thoracotomy chest pain was controlled by a continuous i.v. infusion of a combination of nonnarcotic analgesics, which was progressively switched to an oral administration of analgesics before discharge.
All patients in the study performed a preoperative (usually the day before operation) and an early postoperative (at discharge; mean 10.9 days after operation) PFT. Of these patients, 150 performed DLCO measurements by the single breath method. All spirometry tests were performed according to the American Thoracic Society criteria. Spirometry was performed before and after bronchodilators administration and the best results were used for the present analysis. Thoracotomy chest pain at the time of the postoperative PFT was controlled, if needed, with analgesics administration.
PFT results were expressed as percentages of predicted values for age, gender and height [2]. The postoperative declines of FEV1 and DLCO (FEV1 loss% and DLCO loss%, respectively) were expressed as percentage losses from preoperative values, respectively.
2.1. Statistical analysis
Univariate comparisons between groups was made by means of the Mann–Whitney test, and comparisons between observed and predicted values of FEV1 and DLCO losses were made by means of the Wilcoxon signed rank test.
Stepwise multivariable regression analyses were performed to derive regression equations for the estimation of the decline of FEV1 and DLCO, respectively (dependent variables: FEV1 loss% and DLCO loss%, respectively). The following variables were used as independent variables: oxygen arterial tension (PaO2), carbon dioxide arterial tension (PaCO2), body mass index (BMI), FEV1, forced vital capacity (FVC), FEV1/FVC ratio, DLCO, Chronic Obstructive Pulmonary Disease (COPD)-index [3], percentage of functioning parenchyma to be removed during operation (Func loss%).
Func loss% was estimated by means of CT scan and bronchoscopy findings, and, when used, lung perfusion scan (in patients with a calculated ppoFEV1<50% and in all pneumonectomy candidates [4]).
Multicolinearity among variables was taken into consideration.
Bootstrap resampling procedures with 1000 samples were used to confirm reliability of the variables included in the final models [5].
All statistical tests were 2-tailed, and a significance level of P<0.05 was selected.
The analyses were performed by using the Statview 5.0 (SAS Institute; Cary, NC) and the Stata 8.2 (Stata Corp., College Station, TX) statistical softwares.
3. Results
Table 1 shows the characteristics of the patients included in the study.
Patients submitted to lobectomy had a lower FEV1 loss% (27.2% vs. 33.9, respectively, P=0.001) and DLCO loss% (27% vs. 34.5%, respectively, P=0.09), compared to patients submitted to pneumonectomy.
After L, patients with a preoperative FEV1 <70% of predicted had a lower FEV1 loss% (12.6% vs. 29.8%, respectively, P<0.0001) and DLCO loss% (20.3% vs. 28.3%, respectively, P=0.14), with respect to patients with a preoperative FEV170%.
Among the lobectomy patients with a preoperative FEV1< 70%, 4 of 24 patients (16.7%) improved their postoperative FEV1 and 2 of 21 (9.5%) their postoperative DLCO.
Likewise, patients submitted to L with a preoperative DLCO<70% of predicted had a lower DLCO loss% (20.2% vs. 30.8%, respectively, P=0.004) and FEV1 loss% (21.7% vs. 29.9%, respectively, P=0.003), compared to those patients with a preoperative DLCO70%.
Among the patients with a preoperative DLCO <70%, 7 of 46 (15.2%) improved their DLCO and 5 of 53 (9.4%) improved their FEV1, early after lobectomy.
DLCO corrected for alveolar volume (DLCO/VA) showed an increase after lung resection which was higher after pneumonectomy than after lobectomy (15% vs. 2.6%, respectively, P=0.02).
Multivariable regression analysis yielded the following regression equations (Table 2 and Table 3): estimated FEV1 loss%=21.34 – (0.47xage) + (0.49xFunc loss%) + (17.91xCOPD-index); estimated DLCO loss%=35.99 – (0.31xe) – (36.47xFEV1/FVC ratio) + (0.33xDLCO) + (0.54xFunc loss%).
A stepwise multivariable regression analysis was performed to identify the factors associated with the postoperative increase in DLCO/VA and the only independent significant factor resulted a low preoperative DLCO (regression coefficient –0.28, P=0.003).
There was no difference between observed and estimated (by our regression equations) values of FEV1 loss% (28.2% vs. 28%, respectively, P=0.4) and DLCO loss% (28.1% vs. 28.1%, respectively, P=0.5). Conventional methods of estimation (taking into account the number of functioning segments removed during operation and estimated by means of CT scan, bronchoscopy, and, when used, lung perfusion scan) showed an underestimation of the early FEV1 loss% (predicted 21.4% vs. observed 28.2%, P<0.0001) and DLCO loss% (predicted 21.4% vs. observed 28.1%, P<0.0001). The regression equations did not show significant differences between observed and estimated FEV1 loss% and DLCO loss% even when the results were stratified by the type of operation: a) lobectomy, FEV1 loss%, 27.2% vs. 27%, respectively, P=0.6; DLCO loss%, 27% vs. 27.5%, respectively, P=0.7; b) pneumonectomy, FEV1 loss%, 33.9% vs. 33.8%, respectively, P=0.3; DLCO loss%, 34.5% vs. 31.7%, respectively, P=0.4.
Tables 4 and 5 show the comparisons between observed and predicted (by conventional methods and by our regression equations) values of FEV1 loss% and DLCO loss%, stratified by quartiles of predicted FEV1 loss% and DLCO loss%, respectively. There was a good agreement between the observed and the estimated (by our regression equations) values of FEV1 loss% and DLCO loss%, for each quartile of predicted FEV1 loss% and DLCO loss%, respectively, showing good calibration of the models.
4. Discussion
In addition to the removal of lung tissue, other factors play an important role in the first two weeks after operation in determining the reduction of lung function, such as impairment in chest wall compliance, diaphragm dysfunction, accumulated bronchial secretion, bronchial hyper-reactivity, microatelectasis, increased lung water and reduced surfactant activity [6].
Therefore, the prediction of the residual pulmonary function after lung resection made by the simple calculation of the number of functioning segments removed during operation may underestimate the real functional loss in the early postoperative period. Thus, it was our aim to develop regression equations for the estimation of the early postoperative loss in FEV1 and DLCO to be used for risk stratification purposes.
After lobectomy, patients with a preoperative FEV1 value lower than 70% (moderate to severe COPD, according to the COPD severity classification proposed by the European Respiratory Society) lost approximately 58% less of their FEV1 compared to patients with a preoperative FEV1 70%. This finding demonstrates that the improvement in elastic recoil and airway conductance, which is the base of the minimal deterioration or even the improvement of FEV1 in these COPD patients [3,7–11], may play a role even in the early postoperative phase.
Multiple regression analysis showed the following factors to be significantly associated with the postoperative decline of FEV1: younger age, great percentage of functioning parenchyma removed during operation, and a high COPD-index.
Younger patients have usually a better-preserved lung parenchyma compared to older ones. Therefore, the removal of a similar amount of lung tissue may determine a proportionally greater functional respiratory loss in these patients.
Similarly, the removal of a greater percentage of functioning parenchyma (Func loss%), was predictably associated with a greater loss of FEV1 after operation.
The COPD-index has already been inversely associated with the postoperative reduction in FEV1 [3,7].
Factors significantly associated with the postoperative decline of DLCO were younger age, great Func loss%, high preoperative DLCO value, and a low FEV1/FVC ratio.
In patients with an impaired preoperative diffusion capacity, the tumor-bearing lobe or lung has usually the more damaged parenchyma and the worse ventilation/perfusion mismatch [12]. In these patients, the removal of a similar amount of tissue determines a lower proportional loss in lung diffusion capacity, compared to patients with a higher preoperative DLCO. This was also confirmed by the fact that the only factor associated with the postoperative increase in DLCO/VA in our series was a low preoperative DLCO value.
Patients with lower FEV1/FVC ratio, corresponding to a higher degree of pulmonary obstructive disease, had a greater loss of their DLCO early after lung resection. This may be explained by the fact that patients with more advanced COPD have an already reduced pulmonary vascular bed with a limited capacity of postoperative flow redistribution. Furthermore, they may have a more fragile parenchyma, which may be more easily injured during surgery with resulting areas of edema and contusion that may alter the lung diffusion capacity early after operation. Finally, patients with a low FEV1/FVC ratio are more predisposed to bronchial secretion retention, with intermittent or stable bronchial obstruction and consequent development of areas of air trapping and atelectasis, which in turn lead to V/Q mismatch.
By using our regression equations, the comparison between observed and estimated FEV1 and DLCO losses showed a good agreement, even when the results were stratified by the type of operation performed and by quartiles of predicted FEV1 and DLCO losses, respectively.
In contrast, conventional methods of estimation (by computing the number of functioning segments to be removed by means of CT scan, bronchoscopy and, when used, lung perfusion scan) yielded an underestimation of the respiratory functional loss early after operation.
However, the values of the coefficients of multiple determination (R squared) of the regression models indicate that only part of the variance observed in the early postoperative declines of FEV1 and DLCO may be explained by the independent variables used.
In conclusion, we were able to generate regression models for the estimation of the FEV1 and DLCO losses early after major lung resection in patients with lung cancer. These models showed to be reliable (by bootstrap resampling analyses), well calibrated (by quartiles stratification analyses), and more accurate than conventional methods in the early postoperative phase. However, the proposed models are uniquely meant for risk stratification purposes in the early postoperative phase and should not substitute the traditional ways of prediction of the definitive residual pulmonary function.
Appendix. Conference discussion
Dr. D. Waller (Leicester, UK): It's very interesting and obviously some-thing we have all thought about, how to predict the effects on postoperative function. Could you explain to me how you estimate functional loss again I wasn't clear how you do that.
Dr. Brunelli: The functional loss is basically part of the equation we use traditionally to calculate the ppoFEV1 and ppoDLCO and it is the fraction of functional segments.
Dr. Waller: Estimated how How do you estimate the function
Dr. Brunelli: By means of CT scan or bronchoscoopy. We never use lung perfusion scan. For example, if you have to perform a right upper lobectomy with on obstrucctied segment, you remove 2 functioning segments over 18 functioning segments.
Dr. Waller: Do you use perfusion scans
Dr. Brunelli: We perform perfusion scan only in selected patients, in all pneumonectomy candidates and in those patients with a ppoFEV1 less than 50% of predicted.
Dr. Waller: I would suggest that you can only estimate functional loss from perfusion scanning. Neither CT nor bronschoscopy can assess functional loss. And that is the key to your formula, the estimation of functional loss, but it is not reproducible. That's the problem.
Dr. Brunelli: The functional loss is what we use normally for calculating ppoFEV1. That's just a correction of the traditional equation.
Dr. Dusmet (London, UK): When we study predictions of function postoperatively, I think it is important that no patient be excluded, because each and every one of us is going to be faced with a patient with their chest open and having to make decisions up to and including chest wall resection. So I think any formula that excludes a whole body of patients is, unfortunately, not as useful as we would like.
You just said that you actually had a certain number of your higher risk patients who had an FEV1 that improved after surgery; is that correct
Dr. Brunelli: Yes.
Dr. Dusmet: And what was the predicted value of their postoperative FEV1 using your formula For those patients with improved function, had you predicted an improvement in function using your formula
Dr. Bruenlli: We did not perform this analysis. We didn't check.
Dr. Dusmet: Well, in high-risk patients where you can have a volume-reduction effect, that is the most important piece of information because that's what ultimately decides resectability.
Dr. Brunelli: I can tell you that we had 5 patients with a ppoFEV1 less that 40% of predicted, and these patients had a maean loss of 3% in their FEV1.
Dr. S. Mattioli (Bologna, Italy): When you have an imporvement in FEV1 after lobectomy that means that it was resected tissue affected by sever emphysema. Did you compare your fromula with the CAT scan with density make evaluation of the type of tissue you are going to remove
Dr. Brunelli: We didn't perform a formal radiological/visual assessment in these patients to grace emphysema in their resected lobe. However, what I call tell you is that the COPD indes, for example, in these 24 patients with an FEV1 less that 70% of predicted was 1.15, so they were obstructed clearly with an FEV1/FVC ration of 66% of predicted.
Dr. O. Jorgensen (Odense, Denmark): Maybe I missed the point, but did you validate the model on a new group of patients or did you validate the model on the same group of patients that you used for estimation
Dr. Brunelli: We used bootstrap resampling analysis to validate the model. It is a resampling analysis with replacement. Basically you generate 1,000 new samples from the original data-set with the same number of observations as the original database and then you perform the multiple regression analysis in each of these new samples, and to assess the stability or reliability of the variable, you compare the variables occurring in the final model with those occurring in each of these bootstrap samples. If the variables in the final model occur in more than 50% of the bootstrap models, they can be judged stable or reliable.
Dr. B. Witte (Koblenz, Germany): I would accept your formula in your special clinical setting, but I would not transfer it to any other hospital. Why Because early FEV1 reflects also analgesia access to the thoracic cavity, epidural anesthesia or not, and so on and so on, and I would not transfer this formula to my special clinical setting in my hospital.
Dr. Brunelli: We wanted to have a risk stratification tool more precise than the traditional calculation/estimation of the ppoFEV1 and ppoDLCO, because we observed that in the early phase the loss is greater than what is simply calculated by the removal of lung tissue because a number of other factors play an important role. We are going to use it in the clinical setting, practice.
Footnotes
Presented at the joint 18th Annual Meeting of the European Association for Cardio-thoracic Surgery and the 12th Annual Meeting of the European Society of Thoracic Surgeons, Leipzig, Germany, September 12-15, 2004.
References
Brunelli A, Al Refai M, Monteverde M, Borri A, Salati M, Fianchini A. Stair climbing test predicts cardiopulmonary complications after lung resection. Chest 2002;121:1106–10.
Quanjer PhH, Tammeling GJ, Cotes JE, Pedersen OF, Peslin R, Yernault JC. Lung volumes and forced ventilatory flows. Report Working Party. Standardization of lung function tests. European Community for Steel and Coal. Official statement of the European Respiratory Society. Eur Respir J 1993;6:5–40.
Korst RJ, Ginsberg RJ, Ailawadi M, Bains MS, Downey RJ, Rusch VW, Stover D. Lobectomy improves ventilatory function in selected patients with severe chronic obstructive pulmonary disease. Ann Thorac Surg 1998;66:989–902.
Markos J, Mullan BP, Hillman DR, Musk AW, Antico VF, Lovegrove FT, Carter MJ, Finucane KE. Pre-operative assessment as a predictor of morbidity and mortality after lung resection. Am Rev Respir Dis 1989;139:902–10.
Blackstone EH. Breaking down barriers: helpful breakthrough statistical methods you need to understand better. J Thorac Cardiovasc Surg 2001;122:430–39.
Miyoshi S, Yoshimasu T, Hirai T, Hirai I, Maebeya S, Bessho T, Naito Y. Exercise capacity of thoracotomy patients in the early postoperative period. Chest 2000;118:384–90.
Edwards JG, Duthie DJR, Waller DA. Lobar volume reduction surgery: a method of increasing the lung cancer resection rate in patients with emphysema. Thorax 2001;56:791–95.
Santambrogio L, Nosotti M, Baisi A, Ronzoni G, Bellaviti N, Rosso L. Pulmonary lobectomy for lung cancer: a prospective study to compare patients with forced expiratory volume in 1 s more or less than 80% of predicted. Eur J Cardiothorac Surg 2001;20:684–87.
Carretta A, Zannini P, Puglisi A, Chiesa G, Vanzulli A, Bianchi A, Fumagalli A, Bianco S. Improvement of pulmonary function after lobectomy for non-small cell lung cancer in emphysematous patients. Eur J Cardiothorac Surg 1999;15:602–7.
Sekine Y, Iwata T, Chiyo M, yasufuku K, Motohashi S, Yoshida S, Suzuki M, Iizasa T, Saitoh Y, Fujisawa T. Minimal alteration of pulmonary function after lobectomy in lung cancer patients with chronic obstructive pulmonary disease. Ann Thorac Surg 2003;76:356–62.
Choong CK, Meyers BF, Battafarano RJ, Guthrie TJ, Davids GE, Patterson GA. Lung cancer resection combined with lung volume reduction in patients with severe emphysema. J Thorac Cardiovasc Surg 2004;127:1323–31.
Ali MK, Ewer MS, Atallah MR, Mountain CF, Dixon CL, Johnston DA, Haynie TP. Regional and overall pulmonary function changes in lung cancer. J Thorac Cardiovasc Surg 1983;86:1–8.(Alessandro Brunelli, Arma)
Abstract
The objective of the study was to develop regression models for the prediction of the decline of the forced expiratory volume in one second (FEV1) and the carbon monoxide lung diffusion capacity (DLCO) early after major lung resection. One hundred and ninety patients submitted to pulmonary lobectomy or pneumonectomy for lung cancer performed preoperative and early postoperative (mean 10.9 after operation) pulmonary function tests. One hundred and fifty of these patients also underwent DLCO measurements by the single breath method. The decline of FEV1 and DLCO were expressed as percentage losses from preoperative values. Stepwise multiple regression analyses were performed to develop two models estimating the percent reduction of FEV1 and DLCO from preoperative values. The multivariate procedures were then validated by bootstrap analyses. The following regression equations were derived: estimated percent reduction in FEV1=21.34 – (0.47xage) + (0.49xpercentage of functioning parenchyma removed during operation) + (17.91xCOPD-index); estimated percent reduction in DLCO= 35.99 – (0.31xage) – (36.47xFEV1/FVC ratio) + (0.33xDLCO) + (0.54xpercentage of functioning parenchyma removed during operation). The comparison between observed and estimated losses of FEV1 and DLCO (by using these regression equations) was not significantly different. We think the regression models generated in this study may be reliably used for risk stratification purposes.
Key Words: Lung cancer; Lung resection; Pulmonary function test; Lung diffusion capacity; Postoperative period; Regression analysis
1. Introduction
Major lung resections determine a variable reduction of pulmonary function after operation. Particularly in the early postoperative period, this reduction may be of such extent to cause major morbidity, and hinder the postoperative recovery.
Therefore, the objective of the present study was to develop regression models for the prediction of the decline of two commonly used respiratory parameters, such as the forced expiratory volume in one second (FEV1) and the carbon monoxide lung diffusion capacity (DLCO).
2. Patients and methods
Two hundred and thirty-six patients submitted to lobectomy or pneumonectomy for non small cell lung cancer (NSCLC) from October 2001 through April 2004 were enrolled in our study after giving informed consent. Eight patients undergoing lung resection associated with chest wall resection were excluded. The other 38 patients were excluded from the study because they did not perform the postoperative pulmonary function tests (PFTs) (causes: 10 postoperative deaths, 14 patients with prolonged air leaks discharged with chest tubes and Heimlich valves, 3 patients with empyema and 3 with pneumonia transferred to other local hospitals for further assistance, 2 patients with bronchopleural fistula, 2 patients with postoperative myocardial infarction, 2 patients with postoperative strokes, 1 patient with pulmonary embolism and 1 refusal to perform the PFT).
Thus, the remaining 190 patients (151 males, 39 females; 161 pulmonary lobectomy, L; and 29 pneumonectomy, P) formed the database of the study.
Our criteria for inoperability were a predicted postoperative FEV1 (ppoFEV1) less than 30% of predicted in association with a height climbed at the stair climbing test lower than 12 m, and/or hemodynamic instability [1].
All patients performed lung resection through a muscle-sparing lateral thoracotomy. The following procedures were performed: 40 left upper lobectomies, 39 right upper lobectomies, 37 left lower lobectomies, 24 right lower lobectomies, 4 middle lobectomies, 10 lower bilobectomies, 7 upper bilobectomies, 16 left pneumonectomies, 13 right pneumonectomies.
Postoperative respiratory treatment consisted of daily chest physiotherapy administered by a specialized physiotherapist, early mobilization, incentive spirometer for all patients. Bronchodilators were administered only when clinically indicated. Thoracotomy chest pain was controlled by a continuous i.v. infusion of a combination of nonnarcotic analgesics, which was progressively switched to an oral administration of analgesics before discharge.
All patients in the study performed a preoperative (usually the day before operation) and an early postoperative (at discharge; mean 10.9 days after operation) PFT. Of these patients, 150 performed DLCO measurements by the single breath method. All spirometry tests were performed according to the American Thoracic Society criteria. Spirometry was performed before and after bronchodilators administration and the best results were used for the present analysis. Thoracotomy chest pain at the time of the postoperative PFT was controlled, if needed, with analgesics administration.
PFT results were expressed as percentages of predicted values for age, gender and height [2]. The postoperative declines of FEV1 and DLCO (FEV1 loss% and DLCO loss%, respectively) were expressed as percentage losses from preoperative values, respectively.
2.1. Statistical analysis
Univariate comparisons between groups was made by means of the Mann–Whitney test, and comparisons between observed and predicted values of FEV1 and DLCO losses were made by means of the Wilcoxon signed rank test.
Stepwise multivariable regression analyses were performed to derive regression equations for the estimation of the decline of FEV1 and DLCO, respectively (dependent variables: FEV1 loss% and DLCO loss%, respectively). The following variables were used as independent variables: oxygen arterial tension (PaO2), carbon dioxide arterial tension (PaCO2), body mass index (BMI), FEV1, forced vital capacity (FVC), FEV1/FVC ratio, DLCO, Chronic Obstructive Pulmonary Disease (COPD)-index [3], percentage of functioning parenchyma to be removed during operation (Func loss%).
Func loss% was estimated by means of CT scan and bronchoscopy findings, and, when used, lung perfusion scan (in patients with a calculated ppoFEV1<50% and in all pneumonectomy candidates [4]).
Multicolinearity among variables was taken into consideration.
Bootstrap resampling procedures with 1000 samples were used to confirm reliability of the variables included in the final models [5].
All statistical tests were 2-tailed, and a significance level of P<0.05 was selected.
The analyses were performed by using the Statview 5.0 (SAS Institute; Cary, NC) and the Stata 8.2 (Stata Corp., College Station, TX) statistical softwares.
3. Results
Table 1 shows the characteristics of the patients included in the study.
Patients submitted to lobectomy had a lower FEV1 loss% (27.2% vs. 33.9, respectively, P=0.001) and DLCO loss% (27% vs. 34.5%, respectively, P=0.09), compared to patients submitted to pneumonectomy.
After L, patients with a preoperative FEV1 <70% of predicted had a lower FEV1 loss% (12.6% vs. 29.8%, respectively, P<0.0001) and DLCO loss% (20.3% vs. 28.3%, respectively, P=0.14), with respect to patients with a preoperative FEV170%.
Among the lobectomy patients with a preoperative FEV1< 70%, 4 of 24 patients (16.7%) improved their postoperative FEV1 and 2 of 21 (9.5%) their postoperative DLCO.
Likewise, patients submitted to L with a preoperative DLCO<70% of predicted had a lower DLCO loss% (20.2% vs. 30.8%, respectively, P=0.004) and FEV1 loss% (21.7% vs. 29.9%, respectively, P=0.003), compared to those patients with a preoperative DLCO70%.
Among the patients with a preoperative DLCO <70%, 7 of 46 (15.2%) improved their DLCO and 5 of 53 (9.4%) improved their FEV1, early after lobectomy.
DLCO corrected for alveolar volume (DLCO/VA) showed an increase after lung resection which was higher after pneumonectomy than after lobectomy (15% vs. 2.6%, respectively, P=0.02).
Multivariable regression analysis yielded the following regression equations (Table 2 and Table 3): estimated FEV1 loss%=21.34 – (0.47xage) + (0.49xFunc loss%) + (17.91xCOPD-index); estimated DLCO loss%=35.99 – (0.31xe) – (36.47xFEV1/FVC ratio) + (0.33xDLCO) + (0.54xFunc loss%).
A stepwise multivariable regression analysis was performed to identify the factors associated with the postoperative increase in DLCO/VA and the only independent significant factor resulted a low preoperative DLCO (regression coefficient –0.28, P=0.003).
There was no difference between observed and estimated (by our regression equations) values of FEV1 loss% (28.2% vs. 28%, respectively, P=0.4) and DLCO loss% (28.1% vs. 28.1%, respectively, P=0.5). Conventional methods of estimation (taking into account the number of functioning segments removed during operation and estimated by means of CT scan, bronchoscopy, and, when used, lung perfusion scan) showed an underestimation of the early FEV1 loss% (predicted 21.4% vs. observed 28.2%, P<0.0001) and DLCO loss% (predicted 21.4% vs. observed 28.1%, P<0.0001). The regression equations did not show significant differences between observed and estimated FEV1 loss% and DLCO loss% even when the results were stratified by the type of operation: a) lobectomy, FEV1 loss%, 27.2% vs. 27%, respectively, P=0.6; DLCO loss%, 27% vs. 27.5%, respectively, P=0.7; b) pneumonectomy, FEV1 loss%, 33.9% vs. 33.8%, respectively, P=0.3; DLCO loss%, 34.5% vs. 31.7%, respectively, P=0.4.
Tables 4 and 5 show the comparisons between observed and predicted (by conventional methods and by our regression equations) values of FEV1 loss% and DLCO loss%, stratified by quartiles of predicted FEV1 loss% and DLCO loss%, respectively. There was a good agreement between the observed and the estimated (by our regression equations) values of FEV1 loss% and DLCO loss%, for each quartile of predicted FEV1 loss% and DLCO loss%, respectively, showing good calibration of the models.
4. Discussion
In addition to the removal of lung tissue, other factors play an important role in the first two weeks after operation in determining the reduction of lung function, such as impairment in chest wall compliance, diaphragm dysfunction, accumulated bronchial secretion, bronchial hyper-reactivity, microatelectasis, increased lung water and reduced surfactant activity [6].
Therefore, the prediction of the residual pulmonary function after lung resection made by the simple calculation of the number of functioning segments removed during operation may underestimate the real functional loss in the early postoperative period. Thus, it was our aim to develop regression equations for the estimation of the early postoperative loss in FEV1 and DLCO to be used for risk stratification purposes.
After lobectomy, patients with a preoperative FEV1 value lower than 70% (moderate to severe COPD, according to the COPD severity classification proposed by the European Respiratory Society) lost approximately 58% less of their FEV1 compared to patients with a preoperative FEV1 70%. This finding demonstrates that the improvement in elastic recoil and airway conductance, which is the base of the minimal deterioration or even the improvement of FEV1 in these COPD patients [3,7–11], may play a role even in the early postoperative phase.
Multiple regression analysis showed the following factors to be significantly associated with the postoperative decline of FEV1: younger age, great percentage of functioning parenchyma removed during operation, and a high COPD-index.
Younger patients have usually a better-preserved lung parenchyma compared to older ones. Therefore, the removal of a similar amount of lung tissue may determine a proportionally greater functional respiratory loss in these patients.
Similarly, the removal of a greater percentage of functioning parenchyma (Func loss%), was predictably associated with a greater loss of FEV1 after operation.
The COPD-index has already been inversely associated with the postoperative reduction in FEV1 [3,7].
Factors significantly associated with the postoperative decline of DLCO were younger age, great Func loss%, high preoperative DLCO value, and a low FEV1/FVC ratio.
In patients with an impaired preoperative diffusion capacity, the tumor-bearing lobe or lung has usually the more damaged parenchyma and the worse ventilation/perfusion mismatch [12]. In these patients, the removal of a similar amount of tissue determines a lower proportional loss in lung diffusion capacity, compared to patients with a higher preoperative DLCO. This was also confirmed by the fact that the only factor associated with the postoperative increase in DLCO/VA in our series was a low preoperative DLCO value.
Patients with lower FEV1/FVC ratio, corresponding to a higher degree of pulmonary obstructive disease, had a greater loss of their DLCO early after lung resection. This may be explained by the fact that patients with more advanced COPD have an already reduced pulmonary vascular bed with a limited capacity of postoperative flow redistribution. Furthermore, they may have a more fragile parenchyma, which may be more easily injured during surgery with resulting areas of edema and contusion that may alter the lung diffusion capacity early after operation. Finally, patients with a low FEV1/FVC ratio are more predisposed to bronchial secretion retention, with intermittent or stable bronchial obstruction and consequent development of areas of air trapping and atelectasis, which in turn lead to V/Q mismatch.
By using our regression equations, the comparison between observed and estimated FEV1 and DLCO losses showed a good agreement, even when the results were stratified by the type of operation performed and by quartiles of predicted FEV1 and DLCO losses, respectively.
In contrast, conventional methods of estimation (by computing the number of functioning segments to be removed by means of CT scan, bronchoscopy and, when used, lung perfusion scan) yielded an underestimation of the respiratory functional loss early after operation.
However, the values of the coefficients of multiple determination (R squared) of the regression models indicate that only part of the variance observed in the early postoperative declines of FEV1 and DLCO may be explained by the independent variables used.
In conclusion, we were able to generate regression models for the estimation of the FEV1 and DLCO losses early after major lung resection in patients with lung cancer. These models showed to be reliable (by bootstrap resampling analyses), well calibrated (by quartiles stratification analyses), and more accurate than conventional methods in the early postoperative phase. However, the proposed models are uniquely meant for risk stratification purposes in the early postoperative phase and should not substitute the traditional ways of prediction of the definitive residual pulmonary function.
Appendix. Conference discussion
Dr. D. Waller (Leicester, UK): It's very interesting and obviously some-thing we have all thought about, how to predict the effects on postoperative function. Could you explain to me how you estimate functional loss again I wasn't clear how you do that.
Dr. Brunelli: The functional loss is basically part of the equation we use traditionally to calculate the ppoFEV1 and ppoDLCO and it is the fraction of functional segments.
Dr. Waller: Estimated how How do you estimate the function
Dr. Brunelli: By means of CT scan or bronchoscoopy. We never use lung perfusion scan. For example, if you have to perform a right upper lobectomy with on obstrucctied segment, you remove 2 functioning segments over 18 functioning segments.
Dr. Waller: Do you use perfusion scans
Dr. Brunelli: We perform perfusion scan only in selected patients, in all pneumonectomy candidates and in those patients with a ppoFEV1 less than 50% of predicted.
Dr. Waller: I would suggest that you can only estimate functional loss from perfusion scanning. Neither CT nor bronschoscopy can assess functional loss. And that is the key to your formula, the estimation of functional loss, but it is not reproducible. That's the problem.
Dr. Brunelli: The functional loss is what we use normally for calculating ppoFEV1. That's just a correction of the traditional equation.
Dr. Dusmet (London, UK): When we study predictions of function postoperatively, I think it is important that no patient be excluded, because each and every one of us is going to be faced with a patient with their chest open and having to make decisions up to and including chest wall resection. So I think any formula that excludes a whole body of patients is, unfortunately, not as useful as we would like.
You just said that you actually had a certain number of your higher risk patients who had an FEV1 that improved after surgery; is that correct
Dr. Brunelli: Yes.
Dr. Dusmet: And what was the predicted value of their postoperative FEV1 using your formula For those patients with improved function, had you predicted an improvement in function using your formula
Dr. Bruenlli: We did not perform this analysis. We didn't check.
Dr. Dusmet: Well, in high-risk patients where you can have a volume-reduction effect, that is the most important piece of information because that's what ultimately decides resectability.
Dr. Brunelli: I can tell you that we had 5 patients with a ppoFEV1 less that 40% of predicted, and these patients had a maean loss of 3% in their FEV1.
Dr. S. Mattioli (Bologna, Italy): When you have an imporvement in FEV1 after lobectomy that means that it was resected tissue affected by sever emphysema. Did you compare your fromula with the CAT scan with density make evaluation of the type of tissue you are going to remove
Dr. Brunelli: We didn't perform a formal radiological/visual assessment in these patients to grace emphysema in their resected lobe. However, what I call tell you is that the COPD indes, for example, in these 24 patients with an FEV1 less that 70% of predicted was 1.15, so they were obstructed clearly with an FEV1/FVC ration of 66% of predicted.
Dr. O. Jorgensen (Odense, Denmark): Maybe I missed the point, but did you validate the model on a new group of patients or did you validate the model on the same group of patients that you used for estimation
Dr. Brunelli: We used bootstrap resampling analysis to validate the model. It is a resampling analysis with replacement. Basically you generate 1,000 new samples from the original data-set with the same number of observations as the original database and then you perform the multiple regression analysis in each of these new samples, and to assess the stability or reliability of the variable, you compare the variables occurring in the final model with those occurring in each of these bootstrap samples. If the variables in the final model occur in more than 50% of the bootstrap models, they can be judged stable or reliable.
Dr. B. Witte (Koblenz, Germany): I would accept your formula in your special clinical setting, but I would not transfer it to any other hospital. Why Because early FEV1 reflects also analgesia access to the thoracic cavity, epidural anesthesia or not, and so on and so on, and I would not transfer this formula to my special clinical setting in my hospital.
Dr. Brunelli: We wanted to have a risk stratification tool more precise than the traditional calculation/estimation of the ppoFEV1 and ppoDLCO, because we observed that in the early phase the loss is greater than what is simply calculated by the removal of lung tissue because a number of other factors play an important role. We are going to use it in the clinical setting, practice.
Footnotes
Presented at the joint 18th Annual Meeting of the European Association for Cardio-thoracic Surgery and the 12th Annual Meeting of the European Society of Thoracic Surgeons, Leipzig, Germany, September 12-15, 2004.
References
Brunelli A, Al Refai M, Monteverde M, Borri A, Salati M, Fianchini A. Stair climbing test predicts cardiopulmonary complications after lung resection. Chest 2002;121:1106–10.
Quanjer PhH, Tammeling GJ, Cotes JE, Pedersen OF, Peslin R, Yernault JC. Lung volumes and forced ventilatory flows. Report Working Party. Standardization of lung function tests. European Community for Steel and Coal. Official statement of the European Respiratory Society. Eur Respir J 1993;6:5–40.
Korst RJ, Ginsberg RJ, Ailawadi M, Bains MS, Downey RJ, Rusch VW, Stover D. Lobectomy improves ventilatory function in selected patients with severe chronic obstructive pulmonary disease. Ann Thorac Surg 1998;66:989–902.
Markos J, Mullan BP, Hillman DR, Musk AW, Antico VF, Lovegrove FT, Carter MJ, Finucane KE. Pre-operative assessment as a predictor of morbidity and mortality after lung resection. Am Rev Respir Dis 1989;139:902–10.
Blackstone EH. Breaking down barriers: helpful breakthrough statistical methods you need to understand better. J Thorac Cardiovasc Surg 2001;122:430–39.
Miyoshi S, Yoshimasu T, Hirai T, Hirai I, Maebeya S, Bessho T, Naito Y. Exercise capacity of thoracotomy patients in the early postoperative period. Chest 2000;118:384–90.
Edwards JG, Duthie DJR, Waller DA. Lobar volume reduction surgery: a method of increasing the lung cancer resection rate in patients with emphysema. Thorax 2001;56:791–95.
Santambrogio L, Nosotti M, Baisi A, Ronzoni G, Bellaviti N, Rosso L. Pulmonary lobectomy for lung cancer: a prospective study to compare patients with forced expiratory volume in 1 s more or less than 80% of predicted. Eur J Cardiothorac Surg 2001;20:684–87.
Carretta A, Zannini P, Puglisi A, Chiesa G, Vanzulli A, Bianchi A, Fumagalli A, Bianco S. Improvement of pulmonary function after lobectomy for non-small cell lung cancer in emphysematous patients. Eur J Cardiothorac Surg 1999;15:602–7.
Sekine Y, Iwata T, Chiyo M, yasufuku K, Motohashi S, Yoshida S, Suzuki M, Iizasa T, Saitoh Y, Fujisawa T. Minimal alteration of pulmonary function after lobectomy in lung cancer patients with chronic obstructive pulmonary disease. Ann Thorac Surg 2003;76:356–62.
Choong CK, Meyers BF, Battafarano RJ, Guthrie TJ, Davids GE, Patterson GA. Lung cancer resection combined with lung volume reduction in patients with severe emphysema. J Thorac Cardiovasc Surg 2004;127:1323–31.
Ali MK, Ewer MS, Atallah MR, Mountain CF, Dixon CL, Johnston DA, Haynie TP. Regional and overall pulmonary function changes in lung cancer. J Thorac Cardiovasc Surg 1983;86:1–8.(Alessandro Brunelli, Arma)