INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Lung volume reduction surgery (LVRS) is a palliative treatment for the breathlessness of severe pulmonary emphysema (1). The goal of the operation is to remove 20 to 30% of the lung volume, preferably targeting the regions of most severe emphysema, while preserving lung tissue that is not severely diseased (2).

Because the definition of pulmonary emphysema is based on the anatomic demonstration of destruction of lung tissue, computed tomography (CT) is an imaging modality that is well suited for the in vivo study of emphysema (5). CT detects emphysema with a greater sensitivity than functional tests and can accurately quantify pulmonary emphysema (6- 9). Functional correlates of CT in pulmonary emphysema are well established (FEV₁, FEV₁/FVC ratio, FRC and DL_CO) (10). CT can also estimate accurately the total volume and weight of the two lungs combined, as well as the contraction and expansion of regional lung (8, 11).

In our institution, CT has been used since 1995 in both the preoperative evaluation of candidates for LVRS and in the postoperative assessment of the results. We reviewed our experience retrospectively to compare the accuracy of lung volumes measured by CT and total body plethysmography, to analyze volume changes of individual lungs, to investigate the relation between postoperative improvement in airway obstruction with changes in lung volume and CT densitometry, and to compare the results of bilateral and unilateral LVRS.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Between February, 1995, and June, 1996, 93 patients with diffusely confluent, severe pulmonary emphysema (mean % predicted FEV₁ 23 ± 5%; mean % predicted DL_CO 30 ± 14%) underwent lung volume reduction surgery (LVRS) (1, 2, 14) at our institution. Patients were excluded if they had CT evidence of bronchiectasis or giant bullae. Each of the 93 patients had a history of heavy cigarette smoking. Ventilation/perfusion lung scans were performed in each patient and revealed a heterogeneous distribution of disease, including zones ("target areas") of absent or minimal perfusion, matched by decreased ventilation. CT scans of the chest were obtained for each patient. Twenty-eight patients (14 women, 14 men, 49-75 yr old), each of whom underwent a postoperative CT scan (median 6.7 mo, range 3.2- 8.4 mo) but were otherwise randomly selected, were included in the study. Both the study subgroup and the entire LVRS group each included a single patient with -1-antitrypsin deficiency (also a heavy smoker). Fifteen patients had bilateral LVRS and thirteen patients had unilateral LVRS, which provided a total of 56 individual lungs (43 operated and 13 unoperated) for study.

CT Scans

Twenty-five preoperative and five postoperative CT scans were performed on the same conventional scanner (GE 9800 HiLite; GE Medical Systems, Milwaukee, WI), using contiguous sections, 10-mm collimation and the standard reconstruction algorithm. Three preoperative and 23 postoperative CT scans were performed by the same helical scanner (Somatom 4 Plus; Siemens, Erlangen, Germany) using a pitch of 1, 10-mm collimation, and the standard reconstruction algorithm. Each CT scan included a complete examination of the chest at both TLC and RV. If the patient had difficulty with breath-holding during the acquisition of a helical series, the examination was obtained in two increments. Intermittent rest periods were allowed during conventional scanning to combat patient fatigue. Intravenous contrast medium was not used.

The unwindowed CT images were transferred to a personal computer for an analysis that used custom software written in Visual C++ (Microsoft, Redmond, WA). Each CT scan image was analyzed twice by a seeded, region-growing algorithm (8, 13, 15). In this algorithm, the operator designates a region that consists of a single voxel that is within a lung, then the computer iteratively assesses the adjacent voxels, and those within a range of 500 to 1,024 HU are incorporated into the region. The region continues to grow outward until the point at which no adjacent voxels are within the acceptable range (e.g., all adjacent voxels are chest wall, anterior junction line, bronchus or mediastinum). Occasionally, the boundaries of the trachea, main bronchi and the anterior junction line had to be demarcated manually prior to employing the region-growing algorithm. First, the algorithm was used to isolate the right lung (Figure 1), then repeated to isolate the left lung. After an individual lung was isolated within an image, its volume and the distribution of CT attenuations (Figure 2) were recorded. Following the analysis of all images, the individual lung data were tabulated to give five parameters: individual lung total capacity (TLC_IL), individual lung residual volume (RV_IL), total functional lung volume (TFLV), an emphysema index (EI), and the ratio of the air space volume (V_AS) to the tissue space volume (V_TS).

View larger version (62K): [in this window] [in a new window]   Figure 1.   Computerized segmentation of a CT scan image. Original CT scan image (left panel ). Isolation of the right lung (middle panel ). Separation of the right lung into severely emphysematous and functioning regions using a CT attenuation threshold of 910 HU (right panel ). The fraction of lung volume that is severely emphysematous (denoted in white) is defined to be the EI, while the remaining regions represent the TFLV (see METHODS).

View larger version (21K): [in this window] [in a new window]   Figure 2.   The distribution of CT attenuations for the right lung presented in Figure 1 (center panel ). The vertical dashed line is the 910 HU threshold used to construct the EI and the TFLV (see METHODS). The fraction of pixels with CT attenuations below 910 HU (colored white in the right panel of Figure 1) is defined to be the EI. The area under the curve with attenuations above or equal to 910 HU is defined as the TFLV.

TLC_IL and RV_IL

The air space and tissue space volumes of an individual lung are related to the CT-derived volume of an individual lung (V_IL) by the following relations (16):

(1)

(2)

Equation 1 is based on the assumption that the CT attenuations of air and lung tissue are 1,000 HU and 0 HU, respectively. Errors introduced into V_AS by deviations from these ideal attenuation values are small (unpublished result).

V_AS obtained at TLC is the total capacity for an individual lung (TLC_IL). The sum of the TLC_IL for each of the two lungs corresponds to the TLC. Similarly, V_AS obtained at RV is the RV_IL. The sum of the RV_IL for each of the two lungs corresponds to the RV.

Emphysema Index

Emphysema Index (EI) has been defined as that percentage of a lung's volume that has CT attenuations less than the threshold of 910 HU, which has been shown to be the optimal threshold for 10 mm collimation (5, 13, 15, 17). This parameter is defined only at TLC. The calculation of this parameter is illustrated in Figures 1 and 2.

Air-to-Tissue Ratio

V_AS/V_TS is a measure of overall inflation and is calculated from Equations 1 and 2. Deviations of the air and lung tissue CT attenuations from 1,000 HU and 0 HU, respectively, introduce only small errors into V_AS/V_TS (unpublished results).

TFLV and a CT Scan-Based Prediction of Postoperative FEV₁

Because the EI is that fraction of the lung that is most severely destroyed and poorly functioning, the volume of the remaining lung (with CT attenuations between 500 and 910 HU) represents the "total functional lung volume" (TFLV), i.e., the volume of lung that is relatively less destroyed and therefore most responsible for respiratory function (13, 15). TFLV is defined only at TLC. The calculation of this parameter is illustrated in Figures 1 and 2. A linear relation has been described between the TFLV and the FEV₁ (13). We have used this relation to predict postoperative FEV₁ from CT-calculated preoperative (preop) and postoperative (postop) TFLV, by the following equation:

(3)

In this equation, TFLV is the sum of the TFLV from both lungs because they both contribute to the FEV₁.

Pulmonary Function Tests

Pre- and postoperative (6 mo) spirometric (Warren E. Collins, Inc., Braintree, MA) (FEV₁) and total body plethysmographic (SensorMedics Corp., Yorba Linda, CA) (TLC and RV) studies were available for 26 patients preoperatively and 27 patients postoperatively. The 6-min walk test was also performed and was available for 26 patients preoperatively and 27 patients postoperatively.

Dyspnea Index

Patients subjectively classified their degree of dyspnea pre- and postoperatively (6 mo) according to a 0-5 scale (18).

Data Analysis

Because plethysmography was available for 26 patients preoperatively and 27 patients postoperatively, plethysmography provided a total of 53 TLC and 53 RV measurements (including both pre- and postoperative values). Complete pre- and postoperative CT examinations were obtained on all 28 patients, but the postoperative CT images obtained at RV from two examinations were not available for computer analysis. Thus, CT provided pre- and postoperative TLC_IL for 56 individual lungs and RV_IL for 54 individual lungs. In order to compare CT with plethysmography, we added together either the CT-derived TLC_IL or the RV_IL from the individual right and left lungs to calculate the CT-derived overall TLC and RV, respectively.

CT versus plethysmography plots were constructed for the TLC, RV, VC (assessed as TLC-RV) and RV/TLC ratio, using Excel (Microsoft). Each plot contained both pre- and postoperative measurements. Corresponding pairs of CT and plethysmographic volumes were available for 53 TLC (26 pre- and 27 postoperative) and 49 RV, VC, and RV/TLC ratio (25 pre- and 24 postoperative) measurements.

Statistical Analysis

All statistical calculations were performed using Excel (Microsoft). The comparison of the study subgroup with the entire LVRS group (Table 1) was performed using either unpaired Student's t tests (age, FEV₁, 6-min walk distance, dyspnea index) or ²-tests (gender, type of LVRS). The pre- and postoperative CT parameters in Table 2 were compared using paired Student's t tests. The comparisons of unilateral and bilateral LVRS in Table 3 were performed using unpaired Student's t tests. Pearson's correlation coefficient (r) was used to compare the CT-derived and the plethysmographic lung volumes and the predicted (Equation 3) and measured postoperative FEV₁. The slope and intercept of the regression lines for the CT versus plethysmography plots (Figure 3) were compared with the line of identity by using Student's t tests with the null hypotheses that the slope is equal to 1 and that the intercept is equal to 0. 

View larger version (32K): [in this window] [in a new window]   Figure 3.   Correlation of CT-derived TLC, RV, VC, and RV/TLC ratio with plethysmographic findings. The regression lines are solid, while the line of identity is dashed. Notice that in two cases the CT-derived RV was actually greater than the corresponding CT- derived TLC, which resulted in two CT-derived VC values that were less than zero and two CT-derived RV/TLC values that were greater than one.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Comparision of the Study Subgroup with the Overall LVRS Group

The age and gender distribution, type of LVRS (unilateral or bilateral), pre- and postoperative FEV₁, 6-min walk distance and dyspnea index of the study subgroup were comparable to that of the overall LVRS group (Table 1).

Correlation between Imaging Methods and Plethysmography

TLC and RV calculated from CT correlated well with TLC and RV determined by plethysmography (r = 0.90 [p < 0.0001] and r = 0.84 [p < 0.0001], respectively) (Figure 3). The regression line for TLC, y = 0.91x + 0.17, was not significantly different from the line of identity (intercept: p = 0.72; slope: p = 0.25). The regression line for RV, y = 0.75x + 1.66, was significantly different from the line of identity (intercept: p < 0.0001; slope: p < 0.01). Comparison of the regression line for RV with the line of identity shows CT tended to overestimate RV (Figure 3).

In contradistinction to the TLC and RV measurements, two other CT-derived parameters, the VC and the RV/TLC ratio, correlated significantly but only modestly with plethysmographic results (r = 0.48 [p < 0.0005] and r = 0.33 [p < 0.03], respectively) (Figure 3). The slope of the regression line for VC, y = 0.37x + 0.33, was significantly different from the line of identity, while the intercept was not different (slope: p < 0.0001; intercept: p = 0.18). The intercept of the regression line for RV/TLC, y = 0.30x + 0.62, was significantly different from that of the line of identity, while the slope was not different (intercept: p < 0.0001; slope: p = 0.72). Comparison of the regression lines for VC and RV/TLC with the line of identity shows CT tended to underestimate VC and overestimate RV/TLC (Figure 3).

Pre- and Postoperative Lung Capacity

LVRS significantly reduced the TLC_IL (13%; p < 0.00001) and the RV_IL (20%; p < 0.00001) of operated lungs (Table 2). The absolute and relative (%) magnitude of this reduction was significantly greater at RV_IL than at TLC_IL (p < 0.0008 and p < 0.0001, respectively) (Table 2).

LVRS did not alter the RV_IL of unoperated lungs (Table 2). However, TLC_IL of unoperated lungs increased by 9% (p < 0.0002), so those lungs expanded postoperatively at end-inspiration.

Pre- and Postoperative TFLV and Predicted Postoperative FEV₁

On average, LVRS diminished the operated lung's TLC_IL by 13% (p < 0.00001), but the TFLV actually increased by 9% (p < 0.01) (Table 2). The TFLV for pre- and postoperative scans together with the preoperative FEV₁ allowed the postoperative FEV₁ to be predicted moderately well, using Equation 3 (r = 0.80, p < 0.0001, n = 27; Figure 4).

View larger version (25K): [in this window] [in a new window]   Figure 4.   Correlation of CT-derived predicted postoperative FEV1 with spirometric postoperative FEV1. The CT prediction of postoperative FEV1 is determined by the preoperative FEV1 multiplied by the ratio of postoperative to preoperative total functional lung volume, as determined by CT (METHODS, Equation 3). The regression line is solid, while the line of identity is dashed.

Pre- and Postoperative CT Attenuation and Disease Severity

For individual operated lungs, LVRS decreased significantly the RV_IL/ TLC_IL (10%; p < 0.00001), EI (16%; p < 0.00001), and the V_AS/V_TS at both TLC and RV (14%, p < 0.0002, and 23%, p < 0.00001, respectively) (Table 2). The absolute and relative (%) magnitude of this reduction in the V_AS/V_TS was significantly greater at RV than at TLC (p < 0.0004 and p < 0.002, respectively) (Table 2). In contrast, LVRS did not alter significantly the EI or the V_AS/V_TS of the unoperated lungs (Table 2). However, LVRS significantly decreased the RV_IL/ TLC_IL of unoperated lungs (9%; p < 0.00001).

Pre- and Postoperative FEV₁/VC

By plethysmography, FEV₁/VC was 0.32 ± 0.07 preoperatively and 0.33 ± 0.12 postoperatively (p = 0.87). The FEV₁/ VC (% predicted) was 44 ± 9 preoperatively and 45 ± 14 postoperatively (p = 0.67).

Unilateral Versus Bilateral LVRS

The relative reductions in the TLC_IL, RV_IL, EI, and V_AS/V_TS (at both TLC and RV) for individual operated lungs after either unilateral or bilateral LVRS were not different statistically (Table 3). The absolute and relative (%) magnitude of the reduction in RV_IL was significantly greater than the reduction in TLC_IL for both unilateral and bilateral LVRS (Table 3). The absolute and relative (%) magnitude of the reduction in the V_AS/V_TS was significantly greater at RV than at TLC for both unilateral and bilateral LVRS (Table 3).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

LVRS has a low mortality and morbidity (2, 14, 19, 20) and it frequently offers immediate postoperative relief from disabling respiratory insufficiency. The present study confirms that the beneficial result of surgery is probably due to the synergetic effect of multiple factors (21).

Reduced Air Trapping and Hyperinflation

Prior studies have shown that LVRS significantly reduces hyperinflation of the lung and air trapping, as evidenced by decreases in TLC and FRC (1, 3, 15, 22). The present study shows that LVRS reduced the volume of individual lungs at TLC_IL and at RV_IL (Table 2). Moreover, LVRS reduced air trapping in individual operated lungs, because RV_IL/TLC_IL was reduced postoperatively (Table 2). The fact that LVRS increased the fraction of gas exhaled from individual lungs is further supported by three additional results. First, the absolute and relative (%) magnitude of the average reduction in RV_IL was greater than the reduction in TLC_IL (Table 2). Second, the postoperative reduction in the V_AS/V_TS ratio was greater at RV than it was at TLC (Table 2). Third, because the severity of pulmonary emphysema correlates significantly with the RV/TLC (25), the postoperative reduction in the CT-derived emphysema index (EI) in the present study confirms that LVRS reduced the RV/TLC ratio, and thus reduced air trapping.

After LVRS, the remaining lung parenchyma must stretch to fill the available intrathoracic space. If there were no postoperative remodeling of the chest wall, then pleural pressure would be more negative due to the increased recoil of the remaining lung. Thus, the CT attenuation of at least some lung units can be expected to decrease. It follows that V_AS/V_TS (Equation 1) and EI each can be expected to have a component tending to increase their postoperative values. Conversely, the preferential resection of severely emphysematous tissue can be expected to increase a lung's CT attenuation and to reduce both V_AS/V_TS and EI. The observed postoperative reduction in the V_AS/V_TS and EI (Table 2) shows that the latter factor predominated. The preferential excision of the most severely emphysematous tissues, which tend to retain air during expiration, presumably reduced RV. However, the postoperative increase in TFLV indicates that there was an expansion of lung and perhaps airways in the remaining lung. Furthermore, since TFLV is defined as lung with CT attenuation greater than 910 HU, this postoperative expansion involved the relatively less diseased tissues, which are expected to make an important contribution to the patient's respiratory function.

It must be noted that a reduction in the RV_IL/TLC_IL does not always imply a reduction in the volume of trapped gas. For example, LVRS reduced RV_IL/TLC_IL of the unoperated lungs (Table 2). However, in contrast to the operated lungs, this effect was the result of postoperative increases in TLC_IL, while the RV_IL was unaltered (Table 2).

Reduced Obstruction to Airflow

The original concept of LVRS, as proposed by Brantigan and colleagues (27), was based on the theory that decreased elastic traction on the airways increases airway obstruction and that surgical reduction of lung volume helps restore this elastic traction. Indeed, Gelb and coworkers (22, 26) demonstrated an increase in the slope of the maximum flow-static recoil curve during forced expiration. These findings are consistent with the concept of dilation of airways following LVRS.

We failed to find a change in the FEV₁/VC following LVRS. This result could be interpreted as demonstrating proportional expansion of airways as well as airspace volume in the remaining lung. However, caution is urged in using CT-measured volume data alone to determine the mechanisms limiting maximum expiratory flow in these patients. The situation may be quite complex because a number of competing factors determine the net effect of LVRS on FEV₁. For example, the surgical removal of functioning airways can be expected to decrease FEV₁, while increased recoil can be expected to have the opposite effect.

Unilateral Versus Bilateral LVRS

Comparison of the effect of LVRS on the volume of each lung in the individual patient provides evidence that the beneficial effects of LVRS are due to the surgical alterations in lung morphology, and that these surgical alterations affect each lung independently.

Our results support the results of Cooper and associates who found that the benefit derived from LVRS is in fact due to the operative procedure and cannot be explained solely by the effects of preoperative medical management and exercise rehabilitation (2). After unilateral LVRS, the operated lungs improved their ability to exhale (as evidenced by reductions in lung volume and V_AS/V_TS that were greater at RV than at TLC), while the ability of unoperated lungs to exhale was not altered (Tables 2 and 3). Thus, a patient's enhanced ability to exhale after unilateral LVRS was localized to the operated lung, while the unoperated lung makes no contribution to the postoperative improvement. Moreover, postoperative improvements in FEV₁ correlated with increased TFLV (Figure 4), but TFLV was altered only in operated lungs (Table 2). Therefore, postoperative improvements in FEV₁ after unilateral LVRS appear to localize to the operated lung.

Two aspects of our results indicate that the response of an individual lung to LVRS is independent of the response of the contralateral lung, whether operated or unoperated. First, there was no significant difference between the relative magnitude of the response of individual operated lungs (as measured by postoperative alterations in TLC_IL, RV_IL, EI and V_AS/V_TS) to either unilateral or bilateral LVRS (Table 3). Thus, for example, the postoperative expansion (mean 9% at TLC) of the unoperated lung in unilateral LVRS did not affect the magnitude of the volume reduction in the contralateral lung (Table 2). Second, the reduction in volume and in V_AS/V_TS of the operated lungs was greater at RV than at TLC after either unilateral or bilateral LVRS (Table 3). So, in either instance, the operated lungs were not just smaller, but they were also able to expel a higher fraction of gas than they did preoperatively.

In the case of unilateral LVRS, reduction in the operated lung's volume led to an increase in volume of the contralateral unoperated lung at TLC. On average, the volume expansion of the unoperated lung was approximately 70% of the volume reduction of the operated lung (Table 2). This result suggests the presence of interdependence between the hemithoraces. Presumably, the degree of interdependence is a function of the compliance of the lungs as well as that of the mediastinum. It should be noted that the residual volume of the unoperated lungs was unchanged after unilateral LVRS (Table 2). Thus, although unoperated lungs at TLC_IL were larger postoperatively than preoperatively, the ability of the patient to expel gas from them was not diminished.

Short- and long-term results after bilateral LVRS have been shown to be better than after unilateral LVRS (2, 4, 28). Our data provide imaging evidence that bilateral LVRS is superior to unilateral LVRS simply because it provides a larger volume reduction. If the distribution of emphysema is much more severe in one lung than the other, or if bilateral LVRS is inadvisable, unilateral LVRS remains an effective alternative because it reduces volume and increases the patient's ability to exhale, albeit less than after bilateral LVRS.

In the present series, CT-derived TLC and RV were shown to compare favorably with plethysmographic TLC and RV. This result implies that CT also accurately measured individual lung volumes (TLC_IL and RV_IL). Differences between volumes measured by plethysmography and CT are likely due to the combination of several factors: (1) conventional CT requires repetitive, prolonged breath holding that often fatigues these severely dyspneic patients; (2) variations in the method and vigor with which a patient was coached for the two types of examination; and (3) dependency of lung volume upon position (supine versus sitting) (29). An additional, presumably minor, factor contributing to the discrepancy between lung volumes measured by CT and those measured by plethysmography is that only the latter includes the volume of gas in the trachea and the main bronchi. We believe that the correlation between our CT and plethysmographic findings show that these considerations introduced relatively small errors into the TLC_IL and RV_IL, while the errors in VC and RV_IL/TLC_IL were larger. Also, most of our analysis is based on the relative (%) postoperative alterations in the various CT parameters rather than their absolute values, so systematic errors may tend to cancel out.

A potential limitation of the present study is that two different CT scanners were utilized, so the differences in CT attenuation that we have interpreted as caused by the operative procedure may instead represent technical differences between scanners. We believe, however, that effects, if any, of this difference were minimal because attenuation data obtained by scanners of different leading manufacturers have been shown to be virtually identical (30).