INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Many studies have evaluated criteria for predicting postoperative mortality and cardiopulmonary complications after lung resection. The criteria used to select patients for major pulmonary resections are based on clinical data, spirometry, more detailed pulmonary functional assessment, and cardiac evaluation.

A low preoperative diffusing capacity for carbon monoxide (DL_CO) may identify those patients who have emphysema and a reduced pulmonary capillary bed. They may be prone to postoperative cardiopulmonary complications because of reduction in the pulmonary capillary bed and reduced capacity for gas exchange. The magnitude of DL_CO at rest may not be an adequate reflection of true functional capacity of the lung for gas diffusion, because it may not indicate the capacity of DL_CO to increase during exercise. In addition to the resting DL_CO, the increase in DL_CO with increasing pulmonary blood flow during exercise, indicating recruitment of pulmonary capillary blood, is important in evaluating early abnormalities in diffusing capacity (1). In exercise, both the ventilatory and cardiovascular systems are tested, which can explain observations that exercise variables are predictive of cardiopulmonary complications and mortality. DL_CO increases during exercise, but may not increase adequately if the pulmonary vascular bed is reduced by emphysema.

DL_CO may be overestimated by the conventional single-breath method (SB-DL_CO) because in this method a single equation is used to calculate DL_CO; this equation is valid only for the breathholding phase of the maneuver. In order to avoid problems related to the changing lung volume and timing of the inspiratory and expiratory phases of the SB-DL_CO, Graham and associates (2, 3) used three separate equations (the 3EQ-DL_CO method) to describe CO uptake during each phase of the breathing maneuver: inhalation, breathholding, and exhalation. These equations account analytically for the diffusion of CO during the inhalation, breathholding, and exhalation phases of the single-breath maneuver, eliminating the need to assume that all CO uptake occurs during the breathholding phase. This makes the measurement of SB-DL_CO independent of the breathholding maneuver and increases the precision and accuracy of SB-DL_CO measurement (2) without necessitating a 10-s period of breathholding. Because it obviates the need for breathholding, the 3EQ-DL_CO method is useful in evaluating DL_CO during moderate- to high-intensity exercise, in which prolonged breathholding becomes difficult. We used this method in our study because our patients exercised at moderate intensities and ventilation rates when breathholding became difficult.

The purpose of this prospective study was to evaluate whether abnormal DL_CO is especially useful in predicting postoperative morbidity and mortality after lung resection, and whether lack of an adequate increase in DL_CO during exercise is associated with a greater rate of postoperative complications after lung resection. In this study, we evaluated DL_CO during exercise in patients with lung cancer before lung resection, using the 3EQ-DL_CO method, and related changes in DL_CO during exercise to postoperative complications.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients

All patients with non-small-cell lung cancer scheduled for lung resection at Vancouver General Hospital during the period from October 1998 to May 1999 were evaluated prospectively. We excluded patients who had received radiation treatment or chemotherapy before surgery. The diagnosis and staging of each cancer were based on one or more of the following: chest radiography and computerized tomography (CT); sputum cytology; bronchoscopy with bronchial brushing and/or biopsy; or biopsy or cytology of the lung lesion. The final diagnosis was based on the pathology of the resected lung. Mediastinal involvement was generally excluded by mediastinoscopy or CT scanning of the chest, and metastasis was excluded by CT scanning of brain, liver, and bone if clinically warranted. Patients were excluded from the study if they were over 85 yr old or had symptomatic ischemic heart disease, severe airflow obstruction (FEV₁/FVC < 45%), resting hypoxemia (O₂ saturation [Sa_O2] < 90% at rest), severe restriction in respiratory capacity (FVC < 1.2 L), or another disease that could impair exercise tolerance. After getting permission from their treating surgeons at Vancouver General Hospital, patients who met the selection criteria were informed of the study: (1) directly by their treating surgeons; (2) when they came to the lung function laboratory or preadmission clinic for preoperative evaluation; (3) when they were admitted to the thoracic surgery ward for evaluation prior to surgery; or (4) by mailing of a recruitment notice requesting them to volunteer for the study. Ethical approvals for the study were obtained from the University of British Columbia Clinical Screening Committee for Research and other Studies Involving Human Patients; all patients signed an informed consent form prior to participation.

Lung Function Measurement

Spirometry was done with a computerized dry rolling-seal spirometer (Transfer Test USA [PK Morgan, Chatham, Kent, UK], or Model 922 [Sensormedics, Anaheim, CA]) according to the American Thoracic Society (ATS) criteria (4). The highest values of FEV1 and FVC were taken from the two best tests that agreed to within 5% of each other. Lung volumes were determined with the helium dilution technique on the Morgan spirometer. Patients rebreathed a 10% helium mixture in a closed circuit until equilibration was reached, and the helium concentration was continuously monitored with a helium analyzer. DL_CO was determined by the single-breath technique (5), using an automated valve and timing device and a bag-in-a-box system with the Morgan equipment as previously described (6). The patient inhaled a volume of test gas containing 10% He, 0.3% CO, 21% O₂, and a balance of N₂. A breathholding time of about 10 s was used and was determined by the Jones-Meade method (7). The dead space washout and alveolar sample volumes were set at 900 ml. The determination of DL_CO conformed to ATS standards (8), and DL_CO was corrected for hemoglobin through the use of separate equations for males and females according to ATS recommendations (8). The mean of two DL_CO measurements that were within 5% of each other was taken to be the measured DL_CO, and a maximum of four DL_CO maneuvers, separated from one another by at least 4 min, was used.

3EQ-DL_CO System

The 3EQ-DL_CO system used in our study was similar to the one previously described (2, 3), with slight modifications; a schematic diagram of the system is shown in Figure 1. Patients breathed from a mouthpiece attached to a three-way sliding valve (Model 2870; Hans Rudolph, Kansas City, MO). This could be positioned so that it opened either to a two-way valve (Model 2700; Hans Rudolph) or to the 3EQ-DL_CO system. The two-way valve had inspiratory and expiratory ports to room air. The 3EQ-DL_CO circuit consisted of two one-way valves to separate the inspiratory and expiratory circuits. The one-way valve for the inspiratory circuit led to a nonmixing switching valve, which allowed the patient to inspire either test gas from the inspiratory bag in a sealed Plexiglas box, or room air from the box. The one-way valve for the expiratory circuit emptied into the expiratory bag in the box. The dead space of the equipment was kept to a minimum because an increased dead space will increase the response time of the 3EQ-DL_CO system. A No. 3 Fleisch pneumotachograph (Fleisch, Lausanne, Switzerland) connected to a ± 2 cm H₂O differential pressure transducer (MP45-14-871; Validyne, Northridge, CA) was attached to the box to measure all air flow in and out of the bags or box. Only ambient room air flowed through the pneumotachograph. The pneumotachograph output was amplified in a carrier demodulator (CD15; Validyne) and the flow signal was integrated by the computer software to determine volume. A gas sampling port located just distal to the mouthpiece and connected to a vacuum pump (Model 8805; Sargeant-Welch, Skokie, IL), was used to continuously aspirate gas at a rate of 100 ml/s through the rapidly responding gas analyzers. Gas analyzers employing infrared absorption (BINOS IR Gas Analyzer; Leybold-Heraeus, Hanau, Germany) were used to measure CO and CH₄ levels throughout the 3EQ-DL_CO maneuver. A moderate sampling rate of 100 ml/s produced a 0-to-90% response time of 250 ms with our analyzer. The lag time of the gas analyzer is the transport time entry of gas into the sampling port on the mouthpiece to its entry into the sampling chamber in the gas analyzer. In processing of the data, this value was calculated and added to the response time of the gas analyzers by the computer software. Before daily use of the CO and CH₄ analyzers, their outputs were checked simultaneously for linearity by determining CO and CH₄ concentrations at different dilutions of the inspired gas mixture with room air. It was important to ensure that the system was free from leaks; with low gas concentrations and high aspiration rates, a small leak could make a significant difference in the signal. Water vapor was removed from the sample gas by using Permapure tubing (MD-110-72E; Permapure Inc., Toms River, NJ) that was selectively permeable only to water, and which was kept dry by flushing its exterior with dry O₂ through an external jacket.

View larger version (21K): [in this window] [in a new window]   Figure 1.   Breathing apparatus for the 3EQ-DLCO system. The three-way valve could be switched to the 3EQ-DLCO system or to a two-way valve communicating with room air at both ends.

Patients were seated upright for all 3EQ-DL_CO maneuvers. After being connected into the 3EQ-DL_CO system, patients were asked to follow a previously selected template of the breathing maneuver on a computer monitor in order to guide them through the maneuver (Figure 2). The template had the same shape for all patients, but the relative magnitudes of the inspiratory and expiratory segments were based on the patients' FRC, IC, and VC, which were entered into the computer software (3). Flow rates were set at from 0.5 L/s to 2.5 L/s, depending on each patient's breathing capabilities. The slope of the lines on the breathing-maneuver template allowed the patient to adjust inspiratory and expiratory flow rates and breathholding time. The first part of the breathing maneuver consisted of a deep inspiration of room air from FRC to TLC, a brief period of breathholding, and an expiration to FRC. For the second part of the breathing maneuver, patients inhaled the test gas (0.3% CO, 0.3% CH₄, 21% O₂, and balance N₂) from the inspiratory bag from FRC to TLC, held their breath for about 2 s, and then exhaled to RV.

View larger version (14K): [in this window] [in a new window]   Figure 2.   Breathing template for 3EQ-DLCO maneuver. At the end of tidal breathing, the patient inhaled room air from FRC to TLC with brief breathholding, then exhaled to FRC for the volume history control. The patient then inhaled test gas from FRC to TLC with breathholding for about 2 s before exhaling to RV for the 3EQ-DLCO maneuver. TV = tidal volume.

The 3EQ-DL_CO values were corrected for hemoglobin by using separate equations for males and females according to ATS recommendations (8), as noted earlier. Before the study, the reproducibility of 3EQ-DL_CO measurements at rest and during different levels of steady-state exercise was tested with six normal, healthy volunteers on two separate days. The 3EQ-DL_CO values on the two separate days were 39.91 ± 1.78 ml/min/mm Hg (mean ± SD) at rest and 52.86 ± 2.09 ml/min/mm Hg at 70% of maximal workload. The mean coefficients of variation of the 3EQ-DL_CO at rest and during steady-state exercise were 4.5% and 4.0%, respectively.

Exercise Testing Equipment

Exercise testing was done with a cycle ergometer (Model 800; SensorMedics) and computerized exercise testing equipment (Vmax 229; SensorMedics). Patients wore a nose clip and breathed through a rubber mouthpiece connected to a detachable mass-flow sensor (Vmax/ V6200; SensorMedics) supported by a headpiece. Inspired and expired gases were sampled continuously at the mouth and dried by passage through Permapure tubing before being subjected to automated gas analysis. Oxygen concentrations were measured with a rapidly responding paramagnetic oxygen analyzer, and CO₂ concentrations were measured with an infrared analyzer. The heart rate (HR) and an electrocardiogram (ECG) were recorded continuously by a 12-lead ECG monitor (Model 4000; Quinton, Seattle, WA), and Sa_O2 was monitored with a pulse oximeter (Model S-100e; SiMed, North Bothell, WA). All signals were sampled on a real-time, breath-by-breath basis and stored in a personal computer with a 586 microprocessor. Minute ventilation (E), respiratory rate (RR), O₂, CO₂ production, and respiratory exchange ratio were determined from the volume, O₂, and CO₂ signals, and were shown on-line on the computer monitor during exercise. The exercise equipment was calibrated and its performance was verified daily.

Experimental Protocol

On the first day of the research study, one of us (J.-S.W. or R.T.A.) completed a clinical questionnaire and physical examination of each patient. For the 3EQ-DL_CO procedure the patient rested in a seated position for 15 min to minimize the effect of any prior activity on DL_CO. To familiarize the patient with the breathing template, we conducted several practice runs of the breathing maneuver with the patient breathing room air from the bag-in-box-system; flow rates were adjusted according to the patient's lung function. This process was repeated until the patient's breathing pattern sufficiently matched the flow rates and breathholding time of the breathing template on the computer monitor. Test gas was then introduced into the inspiratory bag, and the resting 3EQ-DL_CO was determined. At least three 3EQ-DL_CO measurements were made, at intervals of at least 5 min; the mean of the three measurements that agreed to within 10% of each other was taken as the resting 3EQ-DL_CO.

Maximal exercise capacity was determined with a stepped incremental exercise test protocol starting at a workload of 15 W, which was increased every minute by 15 W, while a 12-lead ECG, blood pressure, and Sa_O2 were monitored continuously. A vasodilator ointment (Finalgon; 0.4% nonylic acid vanillylamide and 2.5% -butoxyethyl ester of nicotinic acid; Boehringer Ingelheim, Burlington, ON, Canada) was applied sparingly to the ear lobe to improve blood flow, and a pulse oximeter probe was applied to the ear to monitor Sa_O2. If the ear oximeter probe did not provide an accurate reading, a finger probe on the index finger was used. After about 45 s at each workload, patients were asked to indicate their perceived effort for breathing and cycling on the Borg scale (9). The exercise test was discontinued when the patient reached 90% of the maximal predicted HR or felt fatigued and unable to continue, if an abnormal ECG developed, or if the Sa_O2 fell below 85%. The maximal oxygen uptake (O₂max) attained was taken to be the highest O₂ consumption at the highest workload, just before the exercise test was discontinued. The predicted absolute O₂max, expressed per kilogram of body weight, was determined by using the equations of Jones (10). The maximal predicted HR was calculated by subtracting two-thirds of the age of the patient from 210 (10).

Our intent was to make the steady-state exercise 3EQ-DL_CO measurements on our patients on a separate day from that of maximal exercise testing; however, for patients unable to return on this other occasion, we conducted the exercise 3EQ-DL_CO tests on the same day. Patients took a rest for at least 30 min after the progressive exercise test to recover before being subjected to evaluation of 3EQ-DL_CO during steady-state exercise. Testing was done with the patient seated on the cycle ergometer, with monitoring of ECG, blood pressure, and Sa_O2 by a pulse oximetry. After the patient had cycled for 1 min to warm up, the workload was increased to 35% of the premeasured maximal workload and was maintained thus for 3 min, at which point the 3EQ-DL_CO was determined at that workload. The workload was then increased to 70% of the patient's maximal workload for another 3 min at which point the 3EQ-DL_CO at steady-state exercise was determined and was then remeasured after another minute of exercise. Because of the increased ventilation during exercise, it had previously been determined that the minute interval was an adequate time for CO and CH₄ washout.

Postoperative Complications

All thoracotomies and pulmonary resections were performed by two thoracic surgeons (K.G.E. and R.I.F.). The postoperative course of patients was followed carefully, with detailed assessment and recording of complications. Postoperative complications during the patients' hospitalization after resection were classified into mortality, cardiovascular morbidity, and pulmonary morbidity. Cardiovascular morbidity included myocardial infarction (based on symptoms and a change in the ECG or cardiac enzyme elevation), congestive heart failure (requiring therapy), pulmonary edema (chest radiographic evidence), shock (systolic blood pressure < 90 mm Hg), arrhythmia (requiring therapy), and cerebrovascular accident (through brain CT evidence). Pulmonary morbidity included ventilatory support (> 48 h), reintubation, pulmonary embolism (lung scan or pulmonary angiographic evidence), pneumonia (fever, leukocytosis, purulent sputum, and chest radiographic evidence), atelectasis (chest radiographic evidence, and disappearance with respiratory therapy or therapeutic bronchoscopy), and respiratory insufficiency (arterial partial pressure of CO₂ > 45 mm Hg or arterial partial pressure of O₂ < 55 mm Hg during breathing of room air).

Statistical Analysis

Analysis of the study data was done with Microsoft Excel 97 (Microsoft, Seattle, WA) and SPSS version 8.0 (SPSS Inc., Chicago, IL) software on a personal computer 586; we determined the means and SD for the different variables in the overall group of patients and in patients with and without complications. Comparisons of different groups for continuous variables were made with a two-tailed Student's t test. The chi-square test was used for categorical variables. Analysis of multiple variables using stepwise logistic regression (11) was done to investigate the relative usefulness of the combination of different variables for the prediction of postoperative complications. A value of p < 0.05 was considered statistically significant. A receiver operating characteristic (ROC) curve (12) was used to define the best cutoff limits of the different variables in relation to postoperative complications, and sensitivity and specificity were determined for each variable. The area under the ROC curve (AURC) was estimated with the following algorithm: AURC between two successive points = mean sensitivity · difference in (1  specificity).

The total AURC is the sum of successive individual areas. The AURCs were compared through the method of Hanley and McNeil (13).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We attempted to recruit all suitable patients with lung cancer scheduled for thoracotomy and lung resection at Vancouver General Hospital; patients scheduled for minimal invasive surgery (thoracoscopy) were not considered as candidates for our study. Of 88 eligible patients scheduled for lung resection during the period from October 1, 1998 to May 31, 1999, we were able to recruit 65 patients (74%). We compared the 23 patients we could not recruit with the 57 patients we studied, and did not find any significant differences in their preoperative evaluations and postoperative complications. Of the patients recruited for the study, seven were later shown to have advanced cancer by mediastinoscopy, and did not undergo thoracotomy, and one patient refused to have lung resection. The 57 patients we studied were scheduled for lobectomy or more extensive resection, but a total of 13 patients had only segmentectomy, wedge resection, or thoracotomy without resection after surgical exploration.

The 57 patients that we studied had an age of 64 ± 10 yr (mean ± SD); 39 (68%) were men and 18 (32%) were women. Twenty four patients (42%) were smokers, with a mean smoking history of 55 ± 30 pack-years; 22 (39%) were ex-smokers, with a mean smoking history of 34 ± 24 pack-years; and 11 (19%) had never smoked. A total of 14 patients (25%) had sputum production whereas the remaining 43 (75%) did not. Twenty-four patients (41%) had a diagnosis of chronic obstructive pulmonary disease (COPD), 14 (25%) had a diagnosis of hypertension, eight (14%) had a diagnosis of coronary artery disease, and six (11%) had undergone a prior chest operation. Fifty-one patients (89%) fit the criteria for New York Heart Association (NYHA) Class I, 30 (53%) were of performance status 1, 30 (53%) were of exercise capacity 2, and 29 (51%) were of dyspnea scale 1. Laboratory data revealed a mean hemoglobin of 131 ± 17 g/L, albumin of 37 ± 6 g/L, glutamic oxaloacetic transaminase of 26 ± 11 U/L, and creatinine of 87 ± 17 umol/L for the patient group. Chest radiography showed a mass ( 3 cm in diameter) in 29 patients (51%), a nodule in 25 (44%), and consolidation in three (5%); in 30 patients (53%) the lesion was on the right, and in 27 (47%) it was on the left. One patient (2%) had an abnormal ECG (left bundle-branch block) at rest, and six (11%) had an abnormal exercise ECG (left bundle-branch block in one case and ventricular premature contractions in five cases).

The surgical interventions and their numbers performed on the patients performed were 10 pneumonectomies, two bilobectomies, 32 lobectomies, six segmentectomies, four wedge resections, and three thoracotomies without lung resection. In 43 patients (75%), thoracotomy was done through the fifth intercostal space, in one patient it was done through the third intercostal space, in six it was done through the fourth intercostal space, and in seven patients it was done through the six intercostal space. Pericardial dissection was performed in four patients (7%). The patients' mean hospital stay was 10 ± 3 d. Forty-four patients (78%) had primary lung cancer, seven (12%) were found to have benign lesions on final pathologic examination of the resected lung, and six (11%) had metastatic cancer of various types.

Preoperative lung function data are shown in Table 1. The FVC% predicted was 93 ± 16%, but some patients had mild restrictive ventilatory impairment. The FEV₁/FVC was 70 ± 11%, with some patients having some patients having mild or moderate obstructive ventilatory impairment. DL_CO% predicted was 78 ± 19%, with some patients having mild or moderate impairment of diffusing capacity. Preoperative exercise and 3EQ-DL_CO data are shown in Table 2. Maximal exercise capacity was reduced in most patients, with a O₂max% predicted of 66 ± 14%. The lowest Sa_O2 by pulse oximetry was 95 ± 2% (range: 91 to 98%) at rest, and decreased with exercise by a mean of 1 ± 2% (range: 8 to 8%). To adjust for differences in sex, age, and height, 3EQ-DL_CO at rest and exercise was expressed as a percent of the predicted resting SB-DL_CO. The 3EQ-DL_CO at rest (R-DL_CO) was 22.81 ± 8.44 ml/min/mm Hg, and R-DL_CO% predicted was 93 ± 33%. The 3EQ-DL_CO at rest was greater than the SB-DL_CO at rest because inhalation of test gas started from a higher lung volume (14), at FRC in the 3EQ-DL_CO procedure, than that at RV in the SB-DL_CO procedure. DL_CO at 70% of maximal workload (70% DL_CO) was 28.87 ± 10.79 ml/min/mm Hg, and 70% DL_CO% predicted was 119 ± 43%. The increase in 70% DL_CO% predicted from R-DL_CO% predicted ([70%  R]-DL_CO%) was 25 ± 18%, and there was significant interpatient variability in the increase in 3EQ-DL_CO with exercise.

Postoperative complications occurred in 19 patients (33%), and led to death in two (4%); 12 (21%) experienced cardiovascular morbidity; and 13 (23%) had pulmonary morbidity. The cause of the two deaths was pulmonary edema. Arrhythmia (atrial fibrillation in 10 patients and ventricular premature contraction in one patient) was the major cause of cardiovascular morbidity, occurring in 19% of the 57 patients in the study. Two patients had pulmonary edema and one developed shock. Pneumonia was the major cause of pulmonary morbidity, occurring in seven patients (12% of all cases). Five patients had atelectasis, four developed respiratory insufficiency, and two required ventilatory support and reintubation.

Clinical evaluation was compared in patients with complications and those without complications (Table 3). Patients with complications were older than those without complications, and more frequently had diagnosed COPD; they had worse scores on the dyspnea scale, less exercise capacity, and a poorer performance status, but there was no difference in their NYHA classification. Among patients with complications, a higher percentage had sputum production (eight of 19, or 42%) than among patients without complications (six of 38, or 16%), but this difference did not reach statistical significance (p = 0.06). Thoracotomy done through the fifth intercostal space was less frequent in patients with complications, but there were no differences in the frequency of pericardial dissection. All patients with complications had primary lung cancer, whereas six of those without complications had metastatic cancer and seven had benign lesions. As compared with patients without complications, the 19 patients with complications had more extensive lung resection, consisting of six pneumonectomies, one bilobectomy, and 12 lobectomies. The two patients who died had lung resections more extensive than lobectomy, consisting of a pneumonectomy in one case and a bilobectomy in the other. The 12 patients with cardiovascular complications were older than those without cardiovascular complication (71 ± 6 yr versus 62 ± 10 yr, respectively, p < 0.01) and had more extensive lung resection (p < 0.05), consisting of six pneumonectomies, one bilobectomy, and five lobectomies. The 13 patients with pulmonary complications were older than those without pulmonary complications (70 ± 5 yr versus 62 ± 11 yr, respectively, p < 0.01), had lower scores on the dyspnea scale (p < 0.01), less exercise capacity (p < 0.05), and a poorer performance status (p < 0.05), and had more extensive lung resection, consisting of four pneumonectomies, one bilobectomy, and eight lobectomies. A higher percentage of patients with pulmonary complications had sputum production (six of 13) than did those without such complications (eight of 44), but this difference did not reach statistical significance (p = 0.09).

Spirometry was done on all patients, but lung volume measurements and DL_CO were determined in only 47. Lung function tests, including measurement of DL_CO, were compared in patients with and those without complications (Table 4). Patients with complications had lower values of FEV₁% predicted, FVC% predicted, FEV₁/FVC, DL_CO% predicted, and DL_CO/alveolar volume (VA) % predicted. Results of lung function tests were also separately related to the presence or absence of cardiovascular and of pulmonary complications. Patients with cardiovascular complications had lower values of DL_CO% predicted than did patients without cardiovascular complications (60 ± 12% versus 83 ± 17%, respectively, p < 0.001), but there was no significant difference in FEV₁ % predicted. Patients with pulmonary complications had lower values of FEV₁% predicted (68 ± 15% versus 88 ± 18%, respectively, p < 0.001), FEV₁/FVC (61 ± 10% versus 72 ± 10%, respectively, p < 0.01), and DL_CO% predicted (62 ± 14% versus 83 ± 17%, respectively, p < 0.001) than did patients without pulmonary complications.

All patients underwent progressive exercise testing; Table 5 compares results for patients with and without complications. Patients with complications had a lower maximal workload, O₂max, O₂max/kg, and O₂ pulse at maximal workload, indicating moderate impairment of exercise capacity. Analysis of cardiovascular and pulmonary complications done separately showed that patients with cardiovascular complications had a lower O₂max% maximal predicted (57 ± 16% versus 68 ± 13%, respectively, p < 0.05), O₂max/kg (14.7 ± 2.9 ml/ kg/min versus 18.6 ± 4.2 ml/kg/min, respectively, p < 0.01), and O₂ pulse at maximal workload (8.2 ± 2.3 ml/beat versus 10.2 ± 2.6 ml/beat, respectively, p < 0.05) than did patients without cardiovascular complications, indicating greater impairment of exercise capacity. Patients with pulmonary complications had a lower O₂max% maximal predicted (55 ± 14% versus 69 ± 13%, respectively, p < 0.01), and O₂max/kg (14.7 ± 2.0 ml/kg/min versus 18.7 ± 4.4 ml/kg/min, respectively, p < 0.001) than did patients without pulmonary complications, indicating more impairment of exercise capacity.

All patients had their 3EQ-DL_CO determined at rest, whereas DL_CO at 35% of maximal workload was determined in 43 patients and DL_CO at 70% of maximal workload was determined in 55 patients. Table 5 shows that patients with complications had a lower R-DL_CO% predicted, 70% DL_CO% predicted, and (70%  R)-DL_CO% than did patients without complications, indicating impairment of diffusing capacity at rest and an inadequate increase in DL_CO during exercise. The two patients who died had lower values of 70% DL_CO% predicted (75 ± 7% versus 121 ± 44%, respectively, p < 0.01), and (70%  R)-DL_CO% (1 ± 5% versus 26 ± 17%, respectively, p < 0.01) than did the surviving patients, indicating an inadequate increase in DLCO during exercise. It is interesting that the lung function and maximal exercise data of the patients who died were within the range of the surviving patients. Patients with cardiovascular complications had a lower 70% DL_CO% predicted (89 ± 47% versus 127 ± 39%, respectively, p < 0.05) and lower (70%  R)-DL_CO% (6 ± 10% versus 30 ± 17%, respectively, p < 0.001) than did patients without cardiovascular complications, indicating an inadequate increase in DL_CO during exercise. Patients with pulmonary complications had lower values of R-DL_CO% predicted (71 ± 29% versus 100 ± 31%, respectively, p < 0.01), lower 70% DL_CO% predicted (77 ± 29% versus 131 ± 39%, respectively, p < 0.001), and lower (70%  R)-DL_CO% (4 ± 8% versus 31 ± 16%, respectively, p < 0.001) than did patients without pulmonary complications, indicating impairment of diffusing capacity at rest and inadequate increase in DL_CO during exercise.

The analyses presented previously (Tables 345) showed that the four preoperative variables with greatest statistically significant differences (p < 0.001) between patients with and without complications were (70%  R)-DL_CO%, DL_CO% predicted, O₂max/kg, and FEV₁% predicted. Among these variables, (70%  R)-DL_CO% was the best variable in discriminating between patients with and without overall complications, mortality, cardiovascular morbidity, and pulmonary morbidity. Although the four variables showed significant differences between patients with and without overall complications (p < 0.001), only (70%  R)-DL_CO% was significantly different for patients with and without mortality (p < 0.01). For cardiovascular morbidity, (70%  R)-DL_CO%, DL_CO% predicted, and O₂max/kg were significantly different in patients with and without complications but there was no difference in the two groups in FEV₁% predicted. However, all four variables were significantly different in patients with and without pulmonary morbidity (p < 0.001).

Multivariate analysis done through stepwise logistic regression showed that models combining different variables did not significantly improve the prediction of overall complications, mortality, cardiovascular morbidity, or pulmonary morbidity as compared with univariate analysis. Table 6 shows the prediction equations for univariate analysis; the best predictor, as indicated by the most significant p value, appears to be (70%  R)-DL_CO% for overall complications (p < 0.001), cardiovascular morbidity (p < 0.001), and pulmonary morbidity (p < 0.001). For mortality, the prediction equation for (70%  R)- DL_CO% does not reach statistical significance (p = 0.08), owing to the small number of cases (Table 6).

To evaluate the level below which preoperative variables were associated with increased complications, we calculated the incidence of complications for six intervals as follows: ( 25, < 25, < 20, < 15, < 10, and < 5%) for (70%  R)-DL_CO%, ( 24, < 24, < 21, < 18, < 15, and < 12 ml/kg/min) for O₂max/kg, and ( 90, < 90, < 80, < 70, < 60, and < 50%) for DL_CO% predicted and FEV₁% predicted. The ROC curves of these four variables for overall complications, mortality, cardiovascular morbidity, and pulmonary morbidity were determined, and AURC values were calculated. The ROC curves of preoperative variables for prediction of overall complications are shown in Figure 3. The largest AURC was 0.97, calculated from the ROC curve of (70%  R)-DL_CO% for overall complications, and it is statistically significantly greater than that of O₂max/kg, DL_CO% predicted, or FEV₁% predicted (p < 0.001, p < 0.05, and p < 0.001, respectively). This again indicated that (70%  R)-DL_CO% was the best predictor of complications with the best sensitivity and specificity.

View larger version (96K): [in this window] [in a new window]   Figure 3.   ROC curves of preoperative variables for prediction of overall complications. The solid line is the line of identity for a test without any discrimination. (A) ROC curve for (70%  R)- DLCO%; the AURC was 0.97 and the best cutoff point was 10%, with a sensitivity of 78% and specificity of 100%. (B) ROC curve for O2max/kg; the AURC was 0.86 and the best cutoff point was 15 ml/kg/min, with a sensitivity of 58% and specificity of 89%. (C) ROC curve for DLCO% predicted; the AURC was 0.90 and the best cutoff point was 70%, with a sensitivity of 71% and specificity of 93%. (D) ROC curve for FEV1% predicted; the AURC was 0.77 and the best cutoff point was 80%, with a sensitivity of 68% and specificity of 68%.

The best cutoff limit was defined by the point closest to the left upper corner of the ROC curve for each variable, and was 10% in the ROC curve of (70%  R)-DL_CO% for overall complications, with a sensitivity of 78% and specificity of 100% (p < 0.001) (Figure 3A). Fourteen of 18 patients with overall complications had a (70%  R)-DL_CO% of < 10% (sensitivity = 78%), and 37 of 37 patients without complications had a value at or above this limit (specificity = 100%). For O₂max/kg (Figure 3B), a level of 15 ml/kg/min gave the best cutoff limit; 11 of the 19 patients with overall complications had a O₂max/ kg of < 15 ml/kg/min (sensitivity = 58%), and 34 of 38 patients without complications had a O₂max at or above this limit (specificity = 89%). For DL_CO% predicted (Figure 3C), a level of 70% was the best cut-off limit; 12 of 17 patients with overall complications had a DL_CO% predicted of < 70% (sensitivity = 71%), and 28 of 30 patients without complications had a value at or above this limit (specificity = 93%). For FEV₁% predicted (Figure 3D), a level of 80% gave the best cutoff limit; 13 of the 19 patients with overall complications had an FEV₁% predicted of < 80% (sensitivity = 68%), and 26 of 38 patients without complications had a value at or above this limit (specificity = 68%). Furthermore, the various categories of postoperative complications were compared with these cutoff limits for these preoperative lung function variables, as shown in Table 7. Of all the variables, the only predictor of mortality was (70%  R)-DL_CO%. O₂max/kg and DL_CO% predicted did predict overall complications, cardiac morbidity, and pulmonary morbidity, but not mortality, whereas FEV₁ % predicted did predict overall complications and pulmonary morbidity, but not cardiac morbidity or mortality. Thus, the best variable at predicting complications was again (70%  R)-DL_CO%.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The main finding of this study was that patients who developed complications after lung resection had only a limited increase in 3EQ-DL_CO during exercise, from rest to 70% of their maximal workload, expressed as a percent of the predicted resting SB-DL_CO [(70%  R)-DL_CO%], as compared with patients without complications. The results suggested that the best variable in predicting complications was (70%  R)- DL_CO%, as shown in Table 6 and Figure 3. The best cutoff limit of (70%  R)-DL_CO% in predicting complications was < 10% (Table 7 and Figure 3). The limited increase in 3EQ-DL_CO during exercise in patients with complications probably reflects alveolar-capillary membrane destruction and reduction in the pulmonary capillary vascular bed, or limitation of cardiac output. The strong correlation between DL_CO during exercise and postoperative complications was probably due to the contribution of a reduced pulmonary capillary bed to cardiopulmonary complications. DL_CO during exercise appears to be useful as an additional test to improve the prediction of postoperative morbidity after lung resection. This is the first study to show that measurement of DL_CO during exercise, using the three-equation technique, is useful in evaluating patients with lung cancer before lung resection, and in predicting postoperative cardiopulmonary complications.

Wound infection, hematoma, empyema, bronchopleural fistula, air leak for longer than 7 d, and recurrent laryngeal nerve injury were regarded as technical morbidities and were not considered as parts of postoperative cardiopulmonary morbidity in this study. Our retrospective review of complications after pneumonectomy indicated that technical morbidity was not related to patients' lung function, but might have been due to other factors, such as surgical and anesthetic techniques and perioperative care (15). In this prospective study, the postoperative complication rate was 33% and the mortality rate was 4%, resembling those in recent previous reports (16- 18); and the 4% mortality rate was similar to the mortality rate of 5% in our review of 151 pneumonectomy cases (15), and is considered acceptable. Although cardiac arrhythmia was the major cause of morbidity, pulmonary edema was the major cause of mortality in this study and in our retrospective review (15). Pulmonary edema and cardiac dysrhythmias may be induced by the supraphysiologic stresses imposed on the heart and lung during surgery and postoperatively, and by hyperperfusion of the remaining pulmonary vascular bed.

Clinical parameters, including dyspnea scale, exercise capacity, performance status, extent of lung resection, intercostal space used for thoracotomy, and final diagnosis differed significantly for patients with and without cardiopulmonary complications in this study, but showed less significance than age or history of COPD (Table 3). The extent of surgical intervention was also significantly related to mortality as well as to complications; the two patients who died had a pneumonectomy and a bilobectomy, respectively, and all cases with complications had either pneumonectomy or lobectomy. These findings indicate that the amount of lung parenchyma resected was related to the development of postoperative complications including death, and support the conventional approach to using the minimal possible resection. Other clinical factors that may increase the risk of postoperative complications are pulmonary dysfunction, chronic productive cough, cigarette smoking, advanced age, respiratory infections, prolonged anesthesia, and obesity. These clinical factors alter four major aspects of a patient's respiratory status: lung volume, ventilatory pattern, gas exchange, and respiratory defense mechanisms (19).

Lung function testing showed significantly lower values of FEV₁% predicted, FVC% predicted, and FEV₁/FVC in our patients with complications than in those without complications (Table 4). Our analysis showed that spirometry for preoperative evaluation was still a simple and useful predictor of complications. The ability of lung function testing to predict postoperative complications has been variable in previous studies (17, 19, 20). Patient selection, sample size, choice of endpoints, and the retrospective design of some studies are among the possible reasons for the variability in the predictive capacity of lung function testing.

DL_CO% predicted and DL_CO/VA% predicted were significantly lower in patients with complications than in patients without complications in this study (Table 4). A normal DL_CO/VA can be misleading by implying that pulmonary capillary loss in proportion to lung volume loss (i.e., a low DL_CO and low VA with a normal DL_CO/VA) reflects a normal pulmonary capillary bed (21). Thus, DL_CO/VA may not be as good a discriminator as DL_CO in evaluating lung diffusion preoperatively. Both our retrospective study (15) and the present study showed that DL_CO% predicted was better than DL_CO/VA% predicted in its correlation with postoperative complications (Table 4).

Only a few of our patients had quantitative radionuclide lung scanning. However, this technique has been found to be useful in preoperative evaluation (22), especially before pneumonectomy, in estimating the predicted postoperative FEV₁ and DL_CO. One would expect that such a determination, based on the extent of pulmonary resection, preoperative FEV₁, preoperative DL_CO, and proportion of perfusion of the resected lung, would be the preferred method for predicting postoperative lung function. A new index, designated the predicted postoperative product, obtained by multiplying the percent predicted postoperative FEV₁ by the percent predicted postoperative DL_CO, was found by Pierce and colleagues (23) to have the strongest predictive ability for mortality.

Progressive exercise testing showed that maximal workload, O₂max% maximal predicted, O₂max/kg, and O₂ pulse at maximal workload were significantly lower in patients with complications than in patients without complications in this study (Table 5). The purpose of the exercise test is to stress the entire cardiopulmonary oxygen delivery system and estimate the physiologic reserve that may be available after lung resection. During exercise, the lung increases ventilation, O₂ uptake, CO₂ output, and blood flow, simulating increased demand on the remaining lung after lung resection. The viability of the entire system during exercise requires interaction between the lungs, heart, blood vessels, and peripheral muscles.

Resting 3EQ-DL_CO% predicted, 3EQ-DL_CO during exercise at 70% of maximal workload (70% DL_CO% predicted), and the increase in 3EQ-DL_CO with exercise ([70%  R]-DL_CO%) were significantly lower in patients with complications than in patients without complications in this study (Table 5). The strong correlation between diffusing capacity during exercise and postoperative complications is probably due to increased cardiopulmonary complications in patients with a reduced pulmonary capillary bed, smaller amount of available alveolar tissue, and poor recruitment of pulmonary capillary blood volume. Furthermore, 70% DL_CO% predicted and (70%  R)- DL_CO% were significantly lower in patients who died than in patients who did not die in this study. Therefore, (70%  R)- DL_CO% appeared to be a predictor for postoperative mortality in the study, but because of the small number cases, this will require validation in a larger study.

Further analyses, done with logistic regression models (Table 6) and ROC curves with AURC determinations (Figure 3), showed that (70%  R)-DL_CO% was the best predictor of complications, followed in decreasing order by DL_CO% predicted, O₂max/kg, and FEV₁% predicted. Our retrospective review of complications following pneumonectomy also showed that DL_CO% predicted was a better predictor of complications than was FEV₁% predicted (15). A recent study in Chicago also concluded that DL_CO% predicted was a better predictor of complications than was O₂max/kg (24). Another study suggested that the addition of invasive measurement of pulmonary artery pressure during exercise and exercise testing were not helpful in preoperative assessment (25). The index (70%  R)-DL_CO% was derived from the combination of DL_CO measurement and progressive exercise testing, and therefore may be expected to be better than either alone.

The best cutoff limit for (70%  R)-DL_CO% in predicting postoperative complications was 10% (Table 7 and Figure 3), whereas the best cutoff limit for DL_CO% predicted was 70%. The best cutoff limit for O₂max/kg in predicting postoperative complications was 15 ml/kg/min, confirming the findings in previous studies of exercise testing (26, 27). The best cutoff limit for FEV₁ % predicted in predicting postoperative complications was 80%; this high level of prediction by FEV₁ might be due to selection of patients with good values of FEV₁% predicted for lung resection, since FEV₁ was still the main traditionally used test for preoperative evaluation of lung function during the period of selection of our patients for surgery. In addition to the extent of surgical intervention as a risk factor, (70%  R)-DL_CO% was also a significant predictor of postoperative mortality, and the best cutoff limit of (70%  R)-DL_CO% in predicting postoperative mortality was 5%.

Although DL_CO at rest is sensitive enough to detect emphysema, it is not sensitive enough to detect mild emphysema (28). Therefore, in the face of mild disease with a slight reduction in the alveolar capillary surface, the remaining capillaries, with their ability to distend, might be recruited to replace capillaries involved in the emphysematous lesion, yielding a normal value for DL_CO (29). In such cases, measurements of DL_CO during exercise may detect the abnormally reduced alveolar capillary surface and improve the sensitivity of DL_CO for detecting of emphysema, as suggested by Gelb and coworkers (30).

When a portion of the lung is resected, the blood flow to the remaining lung will be increased. The increased blood flow to the remaining pulmonary capillary bed can be accommodated by the recruitment or opening of capillaries that were previously in Zone 1 or Zone 2, or by the distention of capillaries already in Zone 3. Both of these mechanisms can lead to an increase in gas exchange in the remaining lung, compensating to some degree for the loss of the resected portion of the lung. One of the reasons why the preoperative increase in diffusing capacity with exercise is the best predictor of complications from lung resection may be that it measures the ability of the lung to respond to an increase in pulmonary capillary blood flow.

Patients with COPD often complain of exercise intolerance; reduced cardiac output can contribute to limited exercise capacity in patients with severe COPD (31). The tight link between cardiac output and O₂ is usually preserved even in the face of severe COPD; at peak exercise, maximal cardiac output is reduced in patients with COPD as compared with what a normal older patient could achieve at peak exercise, mainly because of ventilatory capacity limiting peak exercise (32). However, cardiac function may be compromised and a higher cardiac output may not be achieved if there is arteriosclerotic heart disease, or if pulmonary hypertension develops during exercise, which may occur in patients with severe COPD (32, 34). Therefore, some patients with emphysema may have a smaller increase in DL_CO during exercise, as a result of decreased recruitment of pulmonary capillary blood volume from a reduced cardiac output during exercise.

Obstructing lesions could reduce DL_CO as well as FEV₁, exercise capacity, and exercise DL_CO. However, obstructing lesions could also result in a shift of blood flow away from the obstructed region of lung to the remaining lung. This could result in increased capillary recruitment in the remaining lung, and an obstructing lesion might therefore affect DL_CO less than FEV₁. If the obstruction were incomplete, DL_CO could be recruited in the obstructed lung during exercise, whereas if the obstruction were complete, DL_CO in the obstructed lung would not be recruited during exercise.

Our study showed that measurement of the increase in DL_CO during exercise, using the 3-equation method and expressed as (70%  R)-DL_CO%, was more useful than measurement of the resting DL_CO, exercise testing, or spirometry in the preoperative evaluation of patients. We are not suggesting that exercise DL_CO be used routinely in the preoperative assessment of patients scheduled for lung resection. In patients with impaired lung function who are at increased risk of complications and who are already undergoing preoperative exercise testing, exercise DL_CO may be of added benefit in evaluating the risk of complications. With the technical progress being made in computerized pulmonary function equipment, and the refinement and availability of rapidly responding gas analyzers for measurement of CO and CH₄, it may be possible in the near future for other centers to determine the 3EQ-DL_CO during exercise for preoperatively evaluating patients scheduled for lung resection and who are undergoing preoperative exercise testing. Measurement of 3EQ-DL_CO during exercise without breathholding is easily accomplished in patients with lung disease; measurement of resting and exercise 3EQ-DL_CO may take additional time, lasting up to 60 min. A prospective study of a larger number of patients will be required to further evaluate the benefit of 3EQ-DL_CO during exercise in the preoperative evaluation of patients for lung resection.