Different Diffusing Capacity of the Lung for Carbon Monoxide as Predictors of Respiratory Morbidity

Different Diffusing Capacity of the Lung for Carbon Monoxide as Predictors of Respiratory Morbidity Robert J. Cerfolio, MD, and Ayesha S. Bryant, MSPH, MD Department of Surgery, Division of Cardiothoracic Surgery, University of Alabama at Birmingham, Birmingham, Alabama Background. The percent predicted diffusing capacity of the lung for carbon monoxide (DLCO%) is an important pulmonary function test (PFT) obtained before elective pulmonary resection. However, there are several DLCO values reported and it is unknown which ones are important predictors of postoperative morbidity. Methods. This is a retrospective study of a prospective database of patients who underwent PFTs and pulmonary resection by one surgeon. The PFTs evaluated were as follows: forced expiratory volume in one second (FEV 1 %), minute ventilation volume (MVV%), and three types of diffusion capacity of the lung for carbon monoxide values: the diffusion capacity of the lung for carbon monoxide (DLCO%), the DLCO adjusted for hemoglobin (DL adjusted%), and the DLCO adjusted for alveolar volume (DLCO/VA%). Results. There were 906 patients between January 2005 and December 2007, and lobectomy was performed most commonly. Complications occurred in 254 patients (28%) and were respiratory in 115 (13%). On univariate analysis, age (p < 0.001), number of cigarettes smoked (p 0.008), history of coronary artery disease (p 0.028), FEV 1 %(p 0.021), postoperative predicted (ppo) FEV1% (p < 0.001), DLCO% (p 0.018), ppodlco% (p 0.002), and DLCO/VA% (p 0.004) were significantly different among those who did and did not experience postoperative respiratory morbidity. Multivariate regression analysis identified ppodlco%, ppofev1%, DLCO/VA%, and age as significant independent predictors of respiratory morbidity. Operative mortality was 2% (18 patients). Conclusions. Although age, FEV 1 %, ppofev 1 %, DLCO%, and ppodlco% are all well-known predictors of operative morbidity after elective pulmonary resection, the DLCO/VA% is another important predictor. This information should be included to help guide patient selection for pulmonary resection and to determine preoperative risk stratification. (Ann Thorac Surg 2009;88:405 11) 2009 by The Society of Thoracic Surgeons The discussion that surgeons, pulmonologists, and oncologists have with patients with nonsmall-cell lung cancer has grown more complicated as the options for therapy have increased. Because surgery continues to offer the optimal survival for patients with early nonsmall-cell lung cancer, this treatment strategy has to be considered in all patients with oncologic favorable disease. However, as our society has grown older and patients more often present with several comorbidities, the risks of all types of therapies have increased. The preoperative identification of patients who are at greater risk is paramount. Pulmonary function testing (PFT) is one of the best preoperative predictors of morbidity and mortality after pulmonary resection. The two most commonly used PFTs to assess patients risk before pulmonary resection are the percent predicted forced expiratory volume in one second (FEV 1 %) and the diffusion capacity of the lung for Accepted for publication April 3, 2009. Presented at the Fifty-fifth Annual Meeting of the Southern Thoracic Surgical Association, Austin, TX, Nov 5 8, 2008. Address correspondence to Dr Cerfolio, Division of Cardiothoracic Surgery, University of Alabama at Birmingham, 703 19th St S, ZRB 739, Birmingham, AL 35294; e-mail: robert.cerfolio@ccc.uab.edu. carbon monoxide (Dlco%). These measurements and their calculated counterparts (the postoperative predictive values) are well known and are complimentary, because the former assesses airflow and the latter assesses gas exchange. However, too many surgeons only consider the FEV 1 % to be the definitive pulmonary function test to perform before lung resection [1]. Many will only perform Dlco% if the FEV 1 % is less than 80%, and some will not even consider the Dlco% regardless of how low the FEV 1 % value may be. The Dlco was initially described in 1950, but it was not widely applied until Ferguson and Vigneswaran [2] and others championed its clinical importance. Ferguson and Vigneswaran [2], Brunelli and colleagues [3], and others [4 6] have shown the clinical efficacy of Dlco% and even the predicted postoperative Dlco% (ppodlco%) [7] in predicting postoperative morbidity and mortality in patients with nonsmall-cell lung cancer and suggest increased risk if the ppodlco% is less than 40%. Most pulmonary function testing centers report more than one Dlco value. The first value often reported is the Dlco% (ml mm Hg 1 min 1 ), and it represents the ability of the lung to diffuse carbon monoxide across its membranes. Unlike the other spirometric measurements, the Dlco is less influenced by patient effort [1]. A second 2009 by The Society of Thoracic Surgeons 0003-4975/09/$36.00 Published by Elsevier Inc doi:10.1016/j.athoracsur.2009.04.015

406 CERFOLIO AND BRYANT Ann Thorac Surg Dlco/VA% AS PREDICTOR OF RESPIRATORY MORBIDITY 2009;88:405 11 value commonly reported is the DL adjusted (ml mm Hg 1 min 1 ) value. This value is the Dlco as described above but it is corrected for the patient s hemoglobin level [8]. The third Dlco value reported is the Dlco/VA (ml mmhg 1 min 1 L 1 ). This represents the Dlco reported above but it is divided by the alveolar volume (VA). Alveolar volume may be a more accurate assessment of functioning alveoli because it is the total lung capacity minus the dead space. The objective of this study was to determine which of these PFTs, and more specifically which Dlco% values, were predictors of postoperative respiratory morbidity after elective pulmonary resection. Patients and Methods Patients This is a retrospective review of a prospective database of a single university-based general thoracic surgeon. Patients who were 19 years of age or older, had PFTs performed within 1 month of surgery, and underwent an elective pulmonary resection were included in this study. Patients who underwent chest wall or diaphragm resection were excluded. Additionally, those who underwent video-assisted thoracoscopic resection were also excluded because over the time of this study, video-assisted thoracoscopic resection was used mainly for diagnosis of interstitial lung disease and not for major pulmonary resection. Patient characteristics we evaluated included age, sex, history of diabetes mellitus, history of hypertension and/or coronary artery disease, performance score (Zubrod score), history of neoadjuvant therapy, type of resection, type of pathology, and side of pulmonary resection. Additionally, preoperative FEV 1 %, minute ventilation volume (MVV%), Dlco%, DL adjusted%, and Dlco/VA% values and their respective ppo values were considered. The University of Alabama at Birmingham s Institutional Review Board approved this study as well as the electronic prospective database used. Individual patient consent was obtained for use in the prospective database but was waived for inclusion in this particular study. Pulmonary Function Test and Dlco Measurement Definitions Pulmonary function tests were performed predominantly at our institution (University of Alabama at Birmingham). Measurements were performed using the Viasys VMAX system (Cardinal Health, San Diego, CA). The Dlco measurement was obtained by having the patient seated upright in a chair with his or her nose pinched closed with a clip. The patient was then asked to breathe normally and then exhale to residual volume. This was a single breath-hold maneuver. At residual volume, a gas mixture (a combination of carbon monoxide and helium) was inhaled forcefully to total lung capacity and held for 10 s and then exhaled. The patient exhaled to wash out an estimate of mechanical and anatomic dead space. An alveolar sample was then collected, and the Dlco was calculated from the total volume of the lung, breath-hold time, and the initial and final alveolar concentrations of carbon monoxide. The exhaled helium concentration was used to calculate a single-breath estimate of total lung capacity and the initial alveolar concentration of carbon monoxide. (The alveolar volume is the total lung capacity minus the physiologic dead space). The driving pressure is assumed to be the calculated initial alveolar pressure of carbon monoxide. The Dlco that is calculated is therefore a product of the patient s single-breath estimate of total lung capacity multiplied by the rate of carbon monoxide uptake during the 10-second breath hold. The average value of two attempts within 10% of each other was recorded as the Dlco. Each PFT variable has multiple values reported: a reference value, a pretherapy (before albuterol is administered) value, and a percent of reference reported. In this study, the percent of reference values will be used consistently for all statistical analysis. The DL-adjusted is based on the patient s measured Dlco% and corrected for the patient s hemoglobin. The calculations used to compute the PFTs are shown below: DLCO VA t2 t1 x FAco (t1) ln FAco (t2) xk Where VA alveolar volume, t2 t1 is time interval over which carbon monoxide (CO) uptake occurred, ln (FAco(t1)/FAco(t2)) is the exponential change in alveolar carbon monoxide concentration over the small time interval (t1 t2), and k is the constant 1,000 60/(PB 47). The Dlco/VA adjusted is simply the Dlco (equation above) divided by the alveolar volume. Alveolar volume (VA) TLC Vd, where TLC is the total lung capacity and the Vd is the physiologic dead space [9]. For patients who underwent lobectomy, bilobectomy, segmentectomy, or wedge resection, ppo values were calculated by the equation: ppofev 1 % measured FEV 1% 100 x 19 # segments resected 19 The numbers of segments resected during surgery were computed as follows: patients who underwent right upper lobectomy were considered to have three segments resected, right middle lobectomy was 2, right lower lobectomy was 5, left upper lobectomy was 5, and left lower lobectomy was 4. If patients underwent a segmentectomy or a wedge, the value 1 was used per wedge or segment. If patients underwent pneumonectomy, the ppo value was based on the preoperative perfusion (V/Q) scan. In this manner, the ppofev1%, ppodlco%, ppodl adjusted%, and ppomvv% values were calculated. Postoperative Definitions Postoperative complications were defined as per the Society of Thoracic Surgeons (STS) database as respiratory (pneumonia, adults respiratory distress syndrome,

Ann Thorac Surg CERFOLIO AND BRYANT 2009;88:405 11 Dlco/VA% AS PREDICTOR OF RESPIRATORY MORBIDITY 407 Fig 1. Patient flow for this study. Of the 3,522 operations performed in this interval, 906 (26%) were eligible for this study. (CW chest wall; PFTs pulmonary function tests; VATS video-assisted thoracoscopy.) reintubation, pneumothorax, air leak longer than 4 days duration, atelectasis requiring bronchoscopy, and bronchopleural fistula). Overall complications were also defined using the STS thoracic data collection form, version 2.07. Prolonged air leak was defined as any air leak persisting past postoperative day 4. Coronary artery disease was defined as any patient who had a history of myocardial infarction, any interventional procedure related to their coronary arteries or a previous coronary Table 1. Patient Characteristics of Entire Cohort (n 906) Characteristic Number of Patients Male 471 (52%) Female 435 (48%) Age mean, median years SD (range) 67, 66 8.8 (20 92) Diabetes mellitus 199 (22%) Hypertension 407 (45%) Coronary artery disease 156 (17%) Zubrod score, mean (range) 1.1 (0 4) Neoadjuvant therapy 163 (18%) Type of resection Pneumonectomy 35 (3%) Lobectomy 548 (60%) Segmentectomy 136 (15%) Wedge 187 (22%) Etiology Nonsmall-cell lung cancer (stages I IV) 652 (72%) Metastatic disease 136 (15%) Benign/infectious etiology 118 (13%) Side of thoracotomy Right 525 (58%) Left 380 (42%) SD standard deviation. artery bypass grafting. Mortality was defined as death before discharge or within 30 days of the operation. Statistical Analysis Analysis was conducted using SAS software 9.1 (SAS Institute, Cary, NC). Continuous data are presented as means (except age, which is presented as median and means) with standard deviations (SD), and categorical data are presented as percentages. Fisher s exact test or Pearson s 2 test were used to assess categorical data. and the Wilcoxon test to evaluate continuous variables. Variables with a p value less than 0.1 at univariate analysis were entered as independent variables in a stepwise logistic regression analysis (dependent variables: respiratory morbidity), with a p 0.05 criterion for retention of variables in the final model. To avoid multicollinearity (between FEV1% and ppofev1%, between Dlco% and ppodlco%, and so forth), only one variable in a set of variables with a correlation coefficient more than 0.5 was Table 2. Postoperative Outcomes of the 906 Patients Included in This Study Morbidity 254 (28%) Respiratory morbidity a 115 (13%) Prolonged air leak 41 Reintubation 9 All other morbidity a 139 (15%) Atrial fibrillation requiring medical therapy 56 Urinary retention 8 Hospital length of stay, days (mean SD) Respiratory morbidity 8.7 5.2 Nonrespiratory/no morbidity 4.1 3.9 Operative mortality 18 (2%) a The two most common morbidities are listed within each category. SD standard deviation.

408 CERFOLIO AND BRYANT Ann Thorac Surg Dlco/VA% AS PREDICTOR OF RESPIRATORY MORBIDITY 2009;88:405 11 Table 3. Univariate Analysis of Patients Who Had a Respiratory Complication (115 Patients) and Those Who Did Not (652 Patients) Respiratory Complications No Respiratory Complications p Value Male 62 (54%) 333 (51%) 0.575 Female 53 (46%) 319 (49%) Age mean, median SD (range) 69, 69 11.5 (24 87) 63, 62 13.0 (20 92) 0.001 Pack-years smoked 47 6.9 33 6.0 0.008 Neo-adjuvant therapy 16 (14%) 127 (19%) 0.194 Diabetes mellitus 24 (21%) 139 (20%) 0.924 Hypertension 35 (30%) 343 (53%) 0.001 Coronary artery disease 12 (10%) 123 (19%) 0.028 Type of resection 0.087 Pneumonectomy 4 (3%) 24 (4%) Lobectomy 77 (67%) 375 (58%) Segmentectomy 19 (17%) 99 (15%) Wedge 15 (13%) 154 (24%) FEV 1 % (L) 70 28.3 79 27.4 0.021 ppofev 1 % 59 29.0 68.8 26 0.001 MVV% a 61 21.6 70 35 0.130 ppomvv% a 51 17.5 56 28.6 0.667 Dlco% (ml mmhg 1 min 1 ) 68 23.6 75 27 0.018 ppodlco% 55 12.3 63 20.1 0.002 DL Adj% (ml mm Hg 1 min 1 ) b 65 15.8 73 29.5 0.176 ppodlco Adj% b 53 12.6 58 13.1 0.287 Dlco/VA% (ml mm Hg 1 min 1 / L 1 ) 88 22.7 96 18.5 0.004 Mean values are reported for continuous variables with SD values. a Reported for only 52% of patients. b Available for 82% of patients. DL Adj% diffusion capacity of lung for carbon monoxide (Dlco) adjusted for hemoglobin; Dlco/VA% Dlco adjusted for alveolar volume; FEV 1 % forced expiratory volume in one second; MVV% minute ventilation volume; ppo postoperative predicted; SD standard deviation. used in the multivariate analysis. The multivariate procedure was validated by bootstrap bagging with 1000 samples [10]. In the bootstrap procedure, repeated samples of 906 observations were selected, with replacement from the original set observations. For each sample, stepwise logistic regression was performed, entering the pulmonary function test and other preoperative variables with p less than 0.1 at univariate analysis. The stability of the final stepwise model can be assessed by identifying the variables that enter most frequently in the repeated bootstrap models and comparing those variables with the variables in the final stepwise model. If the final stepwise model variables occur in a majority (more than 60%) of the bootstrap models, the original final stepwise regression model was judged to be stable. All tests were two-tailed, with a significance level of 0.05. Lastly, a subanalysis was performed to elucidate any clinical significance of the Dlco/VA%, in patients with a normal Dlco% but a depressed Dlco/VA% and vice versa. Results There were 3,522 operations performed between January 2005 and December 2007 by one general thoracic surgeon, and 906 (26%) of these patients (471 men) were eligible for this study. Figure 1 shows the reasons for exclusion. Table 1 depicts the patient characteristics and preoperative PFT values for patients included in this study. The median age was 66 years (range, 20 to 92; mean, 67). As shown in Table 2, respiratory morbidity occurred in 115 patients (13%), and the most common types of respiratory morbidities were prolonged air leak in 41 patients and reintubation secondary to respiratory distress in 9 patients. Nonrespiratory morbidity occurred in 139 patients (15%), and the most common types were atrial fibrillation requiring medical treatment in 56 patients and urinary retention in 8 patients. Twenty-six patients were admitted to the intensive care unit during their hospital stay. The median intensive care unit length of stay was 1.4 days. The mean hospital length of stay for patients with nonrespiratory or no complications was 4.1 days; for patients with respiratory complications, it was 8.7 days (p 0.001). Table 3 depicts the univariate analysis for patients who had respiratory-associated complications. Age at the time of resection, the number of pack-years cigarettes smoked, history of coronary artery disease, FEV 1 %, ppofev 1 %, Dlco%, ppodlco%, and the Dlco/VA% all significantly differed between those who did and did not develop respiratory-associated postoperative morbidity. Logistic regression analysis indicated that significant independent predictors of respiratory morbidity were a reduced ppofev 1 (p 0.001), reduced ppodlco% (p 0.003), reduced Dlco/VA% (p 0.023), greater age (p

Ann Thorac Surg CERFOLIO AND BRYANT 2009;88:405 11 Dlco/VA% AS PREDICTOR OF RESPIRATORY MORBIDITY Table 4. Frequency of Occurrence of Independent Variables in 1,000 Bootstrap Resampling Models Independent Variables 0.001), and smoking history (p 0.045). Table 4 presents the frequency of occurrence of the independent variables in 1,000 bootstrap resampling models: ppodlco% (81.5%), ppofev1% (79.5%), age (70.9%), and Dlco/VA% (67.8%) were the only variables to remain significant in more than 60% of the bootstrap models, indicating their reliability. Operative mortality occurred in 18 patients (2%). The most common causes of mortality were multiorgan system failure initiated by renal failure (7 patients) and reintubation secondary to respiratory distress (3 patients). Twelve of the 18 patients who died were given a do not resuscitate order and died after a prolonged stay in the intensive care unit. In the subanalysis, we found that patients with a low Dlco/VA% but a normal Dlco% had a slightly higher complication rate of 14% (15 of 106 patients) when compared with patients who had a normal Dlco/VA% and a low Dlco% 12% (106 of 130 patients). Comment Frequency of Outcome Age 81.5% ppofev 1 % 79.5% ppodlco% 70.9% Dlco/VA% 67.8% Smoker 9.9% Dlco/VA% Dlco adjusted for alveolar volume; FEV 1 % forced expiratory volume in one second; ppo postoperative predicted. 409 The concept of using a test that measures how well the lungs exchange gas instead of just air flow or the size of the lungs to assess lung function has been studied since the 1950s. The Dlco% assesses the microarchitecture of the alveolus and quantifies the lung s ability to perform its main physiologic function, which is to absorb oxygen and eliminate carbon dioxide. Too often physicians have focused on the volume or size of the lung, and thus on the FEV1% instead of the lung s ability to effectively perform its main function. In fact, too many use the absolute value of the FEV 1 instead of the preferred and more valuable FEV 1 %. We, like most general thoracic surgeons over the past 10 years, have therefore measured the Dlco% and have calculated the ppodlco% on all patients before elective pulmonary resection to gauge their operative risk. It is well known that when either the ppodlco% or the ppofev% are below 40%, the patient is at increased risk. These risk factors along with age, performance status, cardiac function, and exercise capacity have consistently been shown to be accurate predictors of morbidity, not only in patients with chronic obstructive pulmonary disease but also in patients without it [2]. However, the Dlco is reported in three distinct ways by most pulmonary function testing centers, and the clinical significance of these three different values in patients who are to undergo pulmonary resection has not been well studied. Therefore, we performed this study for several reasons. First, to gauge the importance of these three different Dlcos as predictors of respiratory morbidity. We wanted to focus on the less-studied Dlco/VA% and the DL adjusted%, since the prognostic ability of the Dlco% was already well known. Second, we wanted more information on risk assessment. Despite the careful preoperative assessment and the liberal use of pulmonary rehabilitation we try to practice, we continued to observe significant respiratory complications, and some occurred in patients who had a ppodlco% and a ppofev1% above 40% [1, 2, 7]. We wondered if the Dlco/VA% or the DL adjusted% values provided important information or clues about patients at increased risk that we had not previously recognized. We wanted to maximize our preoperative risk stratification as more patients with advanced age and significant comorbidities were presenting for pulmonary resection, and since several new types of less invasive therapies for nonsmall-cell lung cancer are available. On univariate analysis, advance aged, positive smoking history, reduced FEV 1 %, reduced ppofev1%, reduced Dlco%, reduced ppodlco%, and reduced Dlco/VA% were all associated with respiratory morbidity. Surprisingly, we did not find the DL adjusted% to be a predictor, which may be attributed to incomplete/missing data for this variable as it was only available for 82% of the patients in this study. Another surprising finding was that hypertension and coronary artery disease had protective effects with regard to development of respiratory morbidity. Logistic regression analysis, validated by a bootstrap resampling procedure, demonstrated that significant predictors of respiratory morbidity were a reduced ppofev 1, reduced ppodlco%, greater age, and the new finding of this study reduced Dlco/VA%. Thus, the main import of the findings of this study is that not only is ppodlco% a predictor of respiratory complications after pulmonary resection but also so is Dlco/VA%. Both were independent predictors of respiratory-associated complications. Thus, the Dlco/VA% may be another important factor that should be considered in patients who are to undergo pulmonary resection. Of course, it needs to be considered in light of all of the other well-known clinical factors such as age, the ppofev 1 %, and the ppodlco%. The findings in this study make sound clinical sense. The Dlco/VA% measures the Dlco% divided by the effective alveolar volume. Recall that the alveolar volume is the total lung capacity minus the dead space; thus, it represents the effective amount of working lung. Therefore, how much effective alveolar volume that remains after lung resection should be an important predictor of risk, even independent of the FEV 1 % and the Dlco% or their ppo values. In an attempt to further elucidate the clinical significance of the Dlco/VA%, we performed a subanalysis

410 CERFOLIO AND BRYANT Ann Thorac Surg Dlco/VA% AS PREDICTOR OF RESPIRATORY MORBIDITY 2009;88:405 11 which we evaluated patients with a relatively normal Dlco% but a depressed Dlco/VA%. We wondered whether these patients might represent a group of patients who were at an unrecognized increased risk. The results of this analysis were not statistically significant, and the actual difference was very small. We found that patients with a normal Dlco% but a low Dlco/VA% had a slightly higher complication rate when compared with patients who had a low Dlco% but a normal Dlco/VA%: 14% (15 of 106 patients) compared with 12% (106 patients of 130), respectively. Further studies that are powered to detect this difference are needed to see whether these types of patients truly represent a (previously unrecognized) group of patients who are at increased risk after pulmonary resection. There are several limitations to this study. One is that not all PFTs were obtained at the same facility; however, more than 91% were obtained at our institution. When we eliminate the 9% conducted elsewhere, these findings were unchanged. Moreover, previous studies have shown that the Dlco measurements between laboratories are quite (95%) accurate and reproducible if standard techniques are followed [11]. Other potential limitations include patient or instrument error: namely, if the patient is unable to understand the PFT technician, they may exert forces (for example, Valsalva maneuver) during the breath-hold portion of the examination that affects the measurement; measurements may vary owing to miscalibration of the instruments. There are also potential limitations to the statistical methodology we chose to implement to limit interaction and multicollinearity between variables. Only one variable in a set of variables with a correlation coefficient greater than 0.5 was chosen for use in our multivariate analysis to avoid multicollinearity. Selecting the level of the correlation coefficient is arbitrary (for example, it can range from 0.9 to 0.5); however, multicollinearity may occur at even lower levels than 0.5. Second, the choice of which independent variable to remove is relatively arbitrary. Third, multicollinearity can arise if any linear combination of independent variables is correlated with any other linear combination. Thus, examining variables in a pairwise manner will not necessarily remove all sources of multicollinearity. Finally, we did not measure the Dlco% after surgery but rather we used the ppodlco%. Previous series have shown that the ppodlco calculation is an overestimation of the actual decrease in Dlco in the immediate postoperative period [12]. In conclusion, we have shown that not only are well-known values such as patient age, number of cigarettes smoked, FEV 1 %, ppofev 1 %, Dlco%, and the ppodlco% all predictors of respiratory complications, but also, in addition, the Dlco/VA% may be another important factor to consider. This information may help guide patient selection concerning the preoperative risk stratification of pulmonary resection. This study sheds more light on the different types of Dlcos and their clinical significance. Multi-institutional prospective studies are needed to further elaborate these findings and to determine the true clinical significance of this study. References 1. Ferguson MK, Lehman AG, Billiger CT, Brunelli A. The role of diffusing capacity and exercise tests. Thorac Surg Clin 2008;18:9 17. 2. Ferguson MK, Vigneswaran WT. Diffusing capacity predicts morbidity after lung resection in patients without obstructive lung disease. Ann Thorac Surg 2008;85:1158 65. 3. Brunelli A, Refai M, Salati M, Xiume F, Sabbatini A. Predicted versus observed FEV1 and DLCO after major lung resection: a prospective evaluation at different postoperative periods. Ann Thorac Surg 2007;83:1134 9. 4. Boushy SF, Helgason AH, Billig DM, Gyorky FG. Clinical, physiologic, and morphologic examination of the lung in patients with bronchogenic carcinoma and the relation of the findings to postoperative deaths. Am Rev Respir Dis 1960;81:830 8. 5. Markos J, Mullan BP, Hillman DR, et al. Preoperative assessment as a predictor of mortality and morbidity after lung resection. Am Rev Respir Dis 1989;139:902 10. 6. Wang J, Olak J, Ultmann RE, Ferguson MK. Assessment of pulmonary complications after lung resection. Ann Thorac Surg 1999;67:1444 7. 7. Ferguson MK, Reeder LB, Mick R. Optimizing selection of patients for major lung resection. J Thorac Cardiovasc Surg 1995;109:275 283. 8. Levitzky M. Pulmonary physiology. 5th ed. New York: McGraw-Hill, 1999. 9. Goldman H, Becklake M. Respiratory function tests. Am Rev Tuberc 1959;79:457 67. 10. Blackstone EH. Breaking down barriers: helpful breakthrough statistical methods you need to understand better. J Thorac Cardiovasc Surg 2001;122:430 9. 11. American Thoracic Society. Single-breath carbon monoxide diffusing capacity (transfer factor): recommendations for standard technique 1995 update. Am J Respir Crit Care Med 1995;152:2185 98. 12. Varela G, Brunelli A, Rocco G, et al. Predicted versus observed FEV1 in the immediate postoperative period after pulmonary lobectomy. Eur J Cardiothorac Surg 2006;30: 644 8. DISCUSSION DR MARK K. FERGUSON (Chicago, IL): Congratulations, that was a very good presentation. This is a very interesting study that adds a lot of contemporary useful data to the analysis of risk factors for complications after major lung resection. I thought it was interesting that you found that the Dlco corrected for alveolar volume was the strongest predictor of both pulmonary and nonpulmonary complications, and I will touch on that a little bit in a minute. I do have one correction to make. The adjusted Dlco is adjusted for hemoglobin usually, and sometimes for oxygen, but not for height and weight. That comes into the adjustment when you are calculating the percent predicted for all of those variables. The calculation of volume-corrected diffusing capacity can sometimes lead to interesting results. For example, a patient with 60% predicted alveolar volume and 60% predicted Dlco will have a volume-corrected Dlco of 100%. For this reason, my

Ann Thorac Surg CERFOLIO AND BRYANT 2009;88:405 11 Dlco/VA% AS PREDICTOR OF RESPIRATORY MORBIDITY institution elected to stop reporting volume-corrected Dlco many years ago because of the risk of misinterpretation of the reported values. If Dlco is preserved relative to lung volume, then Dlco/VA will reduce apparent risk. In contrast, if Dlco is low relative to lung volume, Dlco/VA will increase apparent risk. This mechanism may help explain why Dlco/VA was a statistically stronger predictor of risk compared to Dlco in your study. Using volume-corrected Dlco for risk assessment could be misleading. What happens if physicians ignore volume measurements and use volume-corrected Dlco as the single predictor of risk? A patient with low Dlco and even lower lung volumes will have a so-called normal volume-corrected Dlco and may be considered suitable for surgery. In terms of your statistical analyses, I would caution you that on the multivariable analyses there is likely to be considerable interaction between the volume measurements and the different types of Dlco that you used concurrently, and so that needs to be looked at very carefully. So with these caveats in mind, I have three questions for you. First, what is the value of including segmental and wedge resection patients in the study? They had half the number of complications compared with major lung resection patients, and this may have diluted the strength of your findings. Second, how was Dlco measurement handled in patients who underwent neoadjuvant therapy, which is known to substantially decrease Dlco, at least temporarily? And finally, how do you explain the fact that both pulmonary and nonpulmonary complications were predicted by Dlco and its variants? Most studies report that things like technical surgical complications and infections are unrelated to physiologic parameters. Again, congratulations on a nice presentation. DR BRYANT: Thank you, Dr Ferguson, for reviewing our paper and for your questions and comments. Your first question centered around including patients who have had lesser resections in this study. Patients with low PFTs often undergo segments, and we wanted to make sure that we included that cohort in this study. Your second question was about how we handle patients who have undergone neoadjuvant therapy, and in those patients, we repeat their PFTs within 1 month before surgery. We have actually found that their PFTs before neoadjuvant therapy and after neoadjuvant therapy have differed significantly, so we started repeating them. And lastly, why was Dlco/VA associated with nonpulmonary complications? I am not really sure why, but perhaps Dlco/VA is some type of a marker in those patients. DR DARRYL S. WEIMAN (Memphis, TN): Any correlation with the stair climb tests with your Dlcos, and also, what nonpulmonary complications were you measuring? DR BRYANT: Thank you, Dr Weiman, for your comments. We did not evaluate whether there was any association with the stair climb test, but that is something that we can go back and look at. Nonpulmonary complications that we were evaluating are all of those that appear on the STS database form, I think it is 7.01, but in our population, the most common ones were atrial fibrillation, myocardial infarction, urinary retention, urinary tract infection, and I think that was it. DR KEITH S. NAUNHEIM (St. Louis, MO): Ayesha, great talk. I know you must be the real brains behind the UAB academic juggernaut because we all know Cerf is a blithering idiot who only provides the muscle. I am wondering if you can give us some advice regarding what you would consider the lower limits of resectability. The lung volume reduction experience has taught us all that we can skate on thinner ice than we ever thought possible, but there still is going to be a point at which we try to do too much and at which patients are too fragile to undergo surgery. I wonder, because you all have had such a large experience over the years, whether or not you can give us your lower limits for an acceptable threshold when you get a predicted postoperative FEV 1 of 32% or 29% or 25%? What is too low? Where should we stop? DR BRYANT: Doctor Cerfolio, do you want to address that? 411 DR CERFOLIO: That is a great question, and I think the answer is different for each patient. Doctor Ferguson s point is critical, and that is, you have got to put the PFTs together with all of the other information you have about the patient. So if you see a 42-year-old man with an FEV 1 of 30% and a Dlco of 30% but he can climb up and down 28 stairs we have a fire escape and use 28 steps I am probably going to operate on him, especially if he doesn t have a cardiac history. Doctor Miller made that point the other day, and I agree, that if their heart is okay, with good RV and LV and no coronary artery disease, that lowers my limit of the Dlco%. And although we probably go overboard, since I get a stress test on almost everybody before surgery and even if they are asymptomatic but if they smoked, they get one we are always evaluating this part. However, if the FEV 1 is low and the Dlco is low and the predicted postoperatives are also low and the patient is elderly, older than 75, and they don t do well on the stair test, I won t. So the truth is, I don t have an absolute value. But to provide an answer to your question given this caveat, I use a PPO FEV 1 of about 30% and a PPO Dlco/VA of about 30% as a marker where I am really worried. What worries me a little bit is I have been using this Dlco/VA because I believe our own research so much, and when you talk to the world s expert, Dr Ferguson, he has really cautioned us on doing that. I agree you must not use this one value alone it has to be placed in context of the FEV 1 % and the other PFTs and their cardiac status and performance status and the stage of the tumor. And I think we do this clinically every day, but I don t think we make a point of that enough in the paper. So I think we are going to have to edit the paper a little before you reviewers chew it up. But that is my short answer.