INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Lung volume reduction surgery (LVRS) improves dyspnea, lung function, and quality of life in selected patients with advanced pulmonary emphysema (1). The improvements are related to restoration of the elastic recoil of the lungs and of the force-generating capacity of the diaphragm (5). According to this concept, selection criteria for LVRS include severe hyperinflation as documented by an increased residual volume (RV) and total lung capacity (TLC), and by signs of increased lung volume in the conventional radiology and the computed tomography (CT) of the chest (1, 9).

In a recent study (12), we found that among patients with a similarly severe degree of hyperinflation and airway obstruction, those in whom preoperative chest CT revealed markedly heterogeneous distribution of lung destruction by emphysema experienced the greatest subjective and functional improvement after LVRS. However, the degree of individual gain varied largely, and even most patients with a diffuse and homogeneous emphysema pattern in the chest CT had significant benefit. In that study (12), regression analysis revealed that 78% of the variation in postoperative gain in FEV₁ could not be explained by preoperative characteristics of pulmonary function or chest CT morphology. Data from other studies confirm that factors not assessed by conventional pulmonary function tests or CT radiography are relevant for surgical outcome (10, 13). Many of these factors have not been clearly identified, and we can only speculate that the amount of reduced volume, the selection of target areas for resection, and the surgical technique are major determinants of outcome. During planning and performing LVRS, the surgeon aims to identify and resect those areas of the lungs that are most destructed by emphysema and, hence, contribute most to hyperinflation and least to ventilation and gas exchange. On the other hand, areas of the lungs with a relatively well maintained function should be preserved. Although chest CT and the video-assisted inspection of the lung during surgery provide an image of lung structure, including the degree and distribution of emphysematous changes, this information does not accurately reflect the function of lung areas targeted for resection. In contrast, lung perfusion scintigraphy, by providing a quantitative display of the regional distribution of one component of lung function, namely blood flow, may identify areas with poor perfusion as targets for resection. Despite the complementary and potentially useful information obtained from perfusion scintigraphy, the role of this technique in the evaluation of LVRS candidates has not been well defined (14, 15).

Therefore, the goal of our study was twofold. First, to design and validate a simple, clinically oriented system for classification of lung-perfusion scans to be applied in LVRS candidates. Second, to correlate the classification of preoperative lung perfusion scans with findings of the chest CT as well as with functional parameters before and 3 mo after surgery. In order to separately estimate the influence of various potential factors that determine outcome after LVRS, we performed multiple regression and contrast analyses among patients grouped according to one factor while controlling for others. In particular, we assessed the effect of emphysema heterogeneity estimated by imaging studies, and that of emphysema severity assumed to be reflected in the reduction of diffusing capacity. Because meaningful comparisons of the predictive value of various tests employed in preoperative evaluation for LVRS are hampered by the confounding effect of the extent of resection, we included a recently proposed measure of the amount of resected lung volume in situ, the fractional reduction in residual volume (1-RV after surgery/RV before surgery [16]) into the model. This analysis should help to more clearly define the role of perfusion scintigraphy, CT imaging, and pulmonary function studies in the evaluation of LVRS candidates.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients

Seventy consecutive patients (24 women) with severe emphysema undergoing LVRS at our institution between August 1994 and November 1997 were included in this study. Their mean age (± SE) was 64 ± 1 yr (range, 43 to 77 yr). All were previous heavy smokers, and eight had homozygous ZZ 1-antitrypsin deficiency. Selection criteria included severe airflow obstruction (FEV₁ < 35% predicted) and hyperinflation (TLC > 130% predicted), dyspnea at rest or on minimal exertion, and smoking cessation. Principal exclusion criteria were hypercapnia (Pa_CO2 > 55 mm Hg), a diffusing capacity for carbon monoxide (DL_CO) < 20% predicted, and significant coronary artery heart disease (17). Patients with bullae 7 cm or more in diameter were also excluded. All patients had been treated with appropriate doses of bronchodilators and had received at least one trial with systemic corticosteroids without success. No systematic rehabilitation was performed before or after the operation.

Measurements and Interventions

Functional evaluation. Maximal flow-volume loops, lung volumes by whole-body plethysmography and DL_CO were measured after inhalation of two puffs of salbutamol (6200 Autobox; SensorMedics, Yorba Linda, CA) according to standard criteria (18, 19) and compared with the reference values of the European Community of Coal and Steel (20). Arterial blood gas analysis was performed while the patient was sitting at rest breathing room air (AVL 995-S; AVL Medical Instruments, Schaffhausen, Switzerland). For assessment of the 6-min walking distance, the patients walked along the same hospital hallway without oxygen supplementation. Dyspnea was rated according to the American Thoracic Society modified Medical Research Council dyspnea score (21), by which shortness of breath is rated with an integer from 0 to 4 according to an increase in symptom severity. Functional evaluation was performed before and 3 mo after the operation.

Imaging techniques. After intravenous application of 150 to 230 mBq ^99mTc-labeled macroaggregates of albumin (SORON Biomedica S.p.A., Vercelli, Saluggia, Italy), lung perfusion scans were obtained in four projections (anterior, posterior, left posterior oblique, and right posterior oblique) with 5 × 10⁵ counts per image each (256 × 256 matrix) on a single-head gamma camera with a digital-integrated system (Diacam; Siemens, Erlangen, Germany). Chest CT examinations were performed on a Somatom plus 4 scanner (Siemens) with high-resolution technique (65 of 70 studies) using an increment of 15 mm and a slice thickness of 1 mm at 140 kV and 11 mA. Scan time for each section did not exceed 1 s.

Surgical technique. Surgical lung volume reduction was performed bilaterally by video-assisted thoracoscopy as described previously (1). The target areas were identified on CT scans and lung perfusion scintigrams and, in some cases, they were selected based on the observation of delayed resorption atelectasis during one lung ventilation. The resection was aimed at the most destroyed tissue using nonbuttressed endoscopic staplers (Endo GIA 30 and 60, Auto-Suture, or ELC 45; Ethicon Endo-Surgery, Cincinnati, OH). A cumulative volume of approximately 20 to 30% of the lung on each side was resected. In cases with no clearly apparent target areas, the resection was performed mostly in the upper lobes.

Data Analysis

Hard copies of the perfusion scintigrams representing the activity counts in the various projections in a gray scale were visually inspected by six observers. Heterogeneity of perfusion was graded according to a semiquantitative staging system that comprised 3 degrees in analogy to a classification recently proposed for chest CT analysis in LVRS candidates (12) (Figure 1).

View larger version (51K): [in this window] [in a new window]   Figure 1.   Perfusion scintigrams were assessed in the posteroanterior (dors), anteroposterior (ventr), and in the left and right posterior oblique views (LPO, RPO). According to the distribution of activity, the set of scans from each patient was assigned to one of three heterogeneity grades. Examples of markedly and intermediately heterogeneous distributions are displayed in panels A and B, and an example of homogeneous distribution is shown in panel C.

Markedly heterogeneous. Distinct regional differences in activity can be identified in at least two adjacent lung segments of either lung.

Intermediately heterogeneous. Distinct regional differences in activity are present maximally in the area of one or more than one, but not in adjacent segments of either lung.

Homogeneous. No or only very minor differences in activity are appreciable indicating fairly homogeneous perfusion of all lung areas. For heterogeneous cases the predominant location of reduced activity (hypoperfusion) was also described as corresponding to either the upper lobes, the upper lobes plus the apical segments of the lower lobes, or the lower lobes.

A visual analog scale for rating scintigrams according to the degree of perfusion heterogeneity was also employed. The examiner had to set a mark onto a straight line printed on a sheet of paper. The relative location of the mark with respect to the two end points of the line, labeled homogeneous and heterogeneous, respectively, represented the degree of heterogeneity. The distance from the point labeled homogeneous to the mark was measured and expressed as percent of the total length of the line.

The scans of the first 48 patients were scored retrospectively by the six observers, who were blinded to the patients names and clinical data. In an initial session, the scoring criteria were defined, and some representative examples for each category were demonstrated to all examiners. Then each of them individually scored the first 20 scans. After the results were recorded, a consensus was obtained for all scans for which scoring was not unanimous. Subsequently, a second set of 28 scans was scored individually by each examiner. After the results were recorded, a consensus was obtained for scans that were not assessed unanimously. Perfusion scintigrams of the last 22 patients in this series were scored before surgery by consensus of at least three observers. For testing intraobserver variability, two investigators reviewed a set of 10 scintigrams a second time 4 wk after the first reading without knowledge of their first assessment.

On the basis of a review of the chest CT, a consensus was obtained among the six observers with respect to scoring of distribution of emphysema according to a classification recently proposed by our group (12). Briefly, emphysematous destruction was described as either markedly heterogeneous (distinct regional differences in destruction in at least two adjacent lung segments of either lung), intermediately heterogeneous (distinct regional differences in destruction but not in adjacent lung segments), or homogeneous (absence of distinct regional differences in destruction). The lobes predominantly affected by emphysema were also determined (upper lobes, upper lobes plus apical segment of lower lobes, or lower lobes). Heterogeneity of emphysema was also rated on a visual analog scale as described above for perfusion scintigrams.

Statistics

Patients were divided into three groups according to the results of perfusion scintigram staging. For each group mean values (± SE) for the results of functional evaluation and chest CT analysis were calculated. Analysis of variance (ANOVA) followed by the Newman-Keuls multiple comparisons procedure when appropriate, was applied to comparisons of preoperative to postoperative values within the same group, and to comparisons between groups. Differences between patients successively grouped according to various potential predictors of outcome after LVRS, while controlling for other factors, were evaluated by defining contrasts within ANOVA. The correlation between changes in pulmonary function and various preoperative baseline characteristics was investigated by multiple linear regression analysis. The performance of several variables in predicting functional outcome after LVRS was further analyzed by plotting receiver operating characteristic (ROC) curves that represent sensitivities and specificities at various cutoff levels (22). The chi-square test was used to compare expected versus observed frequencies. Interobserver and intraobserver variability was assessed by counting the number of observers with identical ordinal rating in the first and second scoring sessions, and the number of identical ratings by the same observers in repeated readings of the same set of scintigrams. The mean discrepancies in visual analog scale ratings (mean of absolute values of differences in ratings, irrespective of algebraic sign) among the six observers and for repeated readings by the same observers were also calculated. A probability of less than 0.05 was considered as statistically significant.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Preoperative Findings

According to the selection criteria for surgery, baseline pulmonary function testing documented severe airflow obstruction and hyperinflation in all patients (Table 1). Arterial blood gas analysis revealed mild-to-moderate hypoxemia, with six patients fulfilling criteria for oxygen therapy (Pa_O2 < 55 mm Hg). The degree of functional impairment and dyspnea was similar among the 10 patients with homogeneous distribution of lung perfusion (mean age, 67 ± 1 yr; six female), 18 patients with intermediately heterogeneous lung perfusion (mean age, 65 ± 2 yr; six female), and 42 patients with markedly heterogeneous lung perfusion (mean age, 63 ± 1 yr, 12 female), respectively (Table 1). In patients with heterogeneous lung perfusion, the areas with the lowest activity were predominantly located in the upper lobes (Table 2).

Interobserver and Intraobserver Variability in Perfusion Scan Readings

Interobserver agreement in assessment of the first 48 perfusion scintigrams was close, with an average of 4.8 identical ordinal scores among the six observers and a mean deviation of visual analog scale scores of 16% (Table 3). Scintigrams with markedly heterogeneous and homogeneous activity distribution were read more consistently than were those with intermediately heterogeneous activity distribution (Table 3). Because the average number of identical readings by the six observers did not differ among the first and the second reading sessions, a relevant learning effect could not be demonstrated (Table 4). Two investigators reanalyzed a subset of 10 perfusion scintigrams after an interval of 4 wk without the knowledge of the results of the first reading. Fourteen of the 20 pairs of ordinal scores were identical and the mean deviation in the visual analog scale scores was 13 ± 3%, indicating a low intraobserver variability.

Postoperative Findings

Three months after LVRS mean values of dyspnea scores and indices of pulmonary function had improved in all three patient groups (Table 1). The gain in lung function measurements listed in Table 1 were more prominent in patients with markedly or intermediately heterogeneous activity in perfusion scintigrams than in patients with homogeneous scintigraphic activity distribution where relatively small changes only in FEV₁ and FVC were noted. However, the intergroup differences in postoperative changes were significant for RV and RV/TLC among patients with markedly heterogeneous versus those with homogeneous scintigrams only (Table 1).

In two of the 10 patients in whom the preoperative perfusion scintigram displayed homogeneous activity distribution, FEV₁ decreased, and in three of 10 there was an increase rather than the expected decrease in the RV/TLC ratio after surgery (Figures 2 and 3). In one of them LVRS had to be interrupted because of cardiorespiratory instability after unilateral operation. A deterioration of hyperinflation was observed in only three of the 60 patients with heterogeneous activity in perfusion scintigrams (Figure 3). There was a small but statistically significant postoperative increase in Pa_O2 in all groups and a decrease in Pa_CO2 in the group with markedly heterogeneous perfusion (Table 1). Three months after the operation only two patients fulfilled criteria for oxygen therapy (Pa_O2 < 55 mm Hg) as compared with six patients preoperatively. Patients with intermediately and markedly heterogeneous lung perfusion experienced an average increase in the 6-min walking distance of approximately one-third (Table 1). In contrast, in one of 10 patients with homogeneous perfusion the 6-min walking distance decreased and the mean value remained unchanged (Table 1). However, the intergroup differences were not statistically significant.

View larger version (20K): [in this window] [in a new window]   Figure 2.   The plots illustrate the effect of thoracoscopic lung volume reduction on FEV1 according to the preoperative heterogeneity of lung perfusion. The latter was graded as either mild (or homogeneous; panel A), intermediately heterogeneous (panel B), or markedly heterogeneous (panel C ). Each panel depicts individual (connected closed circles) and mean (horizontal lines) values before (preop) and after surgery. The p values correspond to comparisons of postoperative to preoperative values. Intergroup comparisons did not reveal significant differences.

View larger version (22K): [in this window] [in a new window]   Figure 3.   The severe degree of hyperinflation was reflected in a high ratio of residual volume to total lung capacity (RV/TLC). This ratio was significantly reduced after surgery in the groups with intermediately and markedly heterogeneous distribution of lung perfusion (panels B and C ). In the group with homogeneous perfusion (panel A), the effect of surgery was not statistically significant, and RV/TLC decreased significantly more in the markedly heterogeneous than in the homogeneous group (p = 0.02).

Subgroup analysis of patients with markedly or intermediately heterogeneous distribution of activity in their perfusion scintigram did not reveal significant differences in changes of pulmonary function parameters or 6-min walking distance among patients with upper lobe predominant hypoperfusion or with left-to-right asymmetry of perfusion compared with patients without such findings.

Correlations of Perfusion Scintigram with Chest CT Readings

The agreement in the degree of lung perfusion heterogeneity as assessed by review of scintigrams with the degree of CT morphologic heterogeneity in distribution of emphysematous lung destruction was only moderate (Table 5). Consistent with this finding, the plots of visual analog scores from perfusion scintigrams versus corresponding chest CT scores revealed a relatively weak correlation (R = 0.51, p < 0.00001) (Figure 4).

View larger version (20K): [in this window] [in a new window]   Figure 4.   Identity plot of visual analog scale scores from chest CTs versus corresponding scores from perfusion scintigrams. The solid line is defined by the regression equation y = 17 + 0.58x (R = 0.51, p < 0.00001). The dotted lines correspond to the 95% confidence intervals of the regression line.

Correlations of Preoperative Imaging and Functional Studies with Outcome

To investigate the independent effect of several potential factors on outcome after LVRS, we performed a series of contrast analyses. Patients were first divided into two groups of equal size (35 patients each) according to whether their scintigraphic visual analog heterogeneity score was greater than the 50th percentile or not. The two groups were further divided into two subgroups each, according to whether DL_CO% pred was greater than the 50th percentile within the respective subgroup or not. Finally, the four resulting subgroups were split further into two subgroups each, according to whether FEV₁% pred was greater than the 50th percentile within the respective subgroup or not. The resulting eight groups (A to H in Table 6) consisting of eight to nine patients were successively arranged into two sets of groups that contrasted with regard to one of three factors (scintigraphic visual analog score, DL_CO% pred, FEV₁% pred) while they were matched for the other two factors. Each set contained three or four groups depending on how matching for all factors that had to be controlled could be achieved. This analysis allowed separate assessment of the effects of the scintigraphic visual analog score, DL_CO% pred and FEV₁% pred (combined with the effect of RV/TLC since they were closely correlated) on several outcome variables (changes in FVC, FEV₁, RV/TLC, and RV). The statistical analysis (Table 6) revealed significant effects of DL_CO% pred on the postoperative gain in FEV₁ and FVC (lower values of preoperative DL_CO were associated with greater gains in these variables), and a significant combined effect of FEV₁ and RV/TLC on changes in FEV₁. The scintigraphic visual analog score did not show a significant effect on any of the outcome variables.

Another contrast analysis was performed by first dividing the patients into two groups of equal size according to whether their chest CT visual analog heterogeneity score was greater than the 50th percentile or not. Subsequently, subgroups were built according to the individual values of DL_CO and FEV₁ as described above for the first contrast analysis. Sets of the resulting eight groups (A to H in Table 7) were arranged to successively obtain contrasts for chest CT visual analog scores, DL_CO, and FEV₁ while matching for the two remaining variables. Significant effects were found for DL_CO on changes in FEV₁, and for FEV₁ and RV/TLC (combined) on changes in FVC, whereas there was no effect of chest-CT visual analog scores on outcome variables.

Multiple linear regression analysis between several baseline characteristics and the three outcome variables FVC%, FEV₁%, and RV/TLC revealed significant correlations between preoperative RV/TLC and changes in FVC and RV/ TLC, between chest CT visual analog scores and changes in FVC, RV/TLC, and between preoperative FEV₁ and changes in FVC (Table 8). Scintigraphic visual analog scores were not correlated with any of the three outcome variables (Table 8). If the relative change in RV (RV% = [RV after surgery  RV before surgery]/RV before surgery expressed in percent = the negative value of 1-RV after surgery/RV before surgery) was included into the model as a measure of resected lung volume, then the multiple R was significantly increased to R = 0.72 (p < 0.0001) for FVC, to R = 0.64 (p < 0.0001) for FEV₁, and to R = 0.81 (p < 0.0001) for RV/TLC. The significant correlation among the chest CT visual analog score and RV% (R = 0.35, p = 0.003) suggested an interaction among them.

There was a significant negative correlation between RV% with DL_CO% pred (R = 0.29, p < 0.0001). In other words, the greater the estimated amount of resected lung volume (1-RV after surgery/RV before surgery = the negative value of RV%) the greater the gain in diffusing capacity. Baseline DL_CO% pred was negatively correlated with changes in DL_CO (DL_CO% pred) (R = 0.67, p < 0.0001). Because mean DL_CO remained unchanged, the latter indicated that patients with a lower diffusing capacity before surgery tended to experience an increase of DL_CO, whereas those with a higher preoperative DL_CO tended to have a decrease. In fact, none of the nine patients with a preoperative value of DL_CO < 30% pred had a decrease in diffusing capacity.

To investigate the ability of certain preoperative baseline studies to predict a minimal postoperative gain in FEV₁ of > 0.3 L or in FVC of > 1 L in individual patients, receiver operating characteristic (ROC) curves were plotted. The levels of the criteria were arbitrarily selected to clearly exceed the measurement variability. The analysis was applied to scintigraphic and chest CT visual analog heterogeneity scores, RV/ TLC, and DL_CO% pred. Of the two curves for each of the four parameters, the one with the larger area is displayed in Figure 5. The remaining areas (± SE) were 0.66 ± 0.07 (p = 0.006) for DL_CO% pred, criterion FEV₁ > 0.3 L; 0.63 ± 0.07 (p = 0.03) for chest CT visual analog heterogeneity scores, criterion FVC > 1 L; 0.51 ± 0.06 (p = NS) for RV/TLC, criterion FEV₁ > 0.3 L; and 0.59 ± 0.07 (p = NS) for scintigraphic visual analog heterogeneity scores, criterion FEV₁ > 0.3 L. The cutoff values corresponding to a specificity of  95% for areas significantly greater than 0.5 for the criterion FEV₁ > 0.3 L were:  89% for chest CT visual analog heterogeneity scores and  23% for DL_CO% pred. The cutoff values corresponding to a specificity of  95% for areas significantly greater than 0.5 for the criterion FVC > 1 L were:  89% and  92% for chest CT and scintigraphic visual analog heterogeneity scores, respectively,  77% for RV/TLC and  27% for DL_CO% pred. Of note, because of the negative correlation of DL_CO with gains in FEV₁ and FVC (Table 8), observed values of DL_CO had to be equal or less (rather than greater) than the cutoff levels to be counted as positive (suggesting that the expected gain in outcome variables would be achieved).

View larger version (33K): [in this window] [in a new window]   Figure 5.   Receiver operating characteristic curves for the criterion "gain in FEV1 > 0.3 L" for chest CT visual analog heterogeneity scores (panel A), and for the criterion "gain in FVC > 1 L" for scintigraphic visual analog heterogeneity scores (panel B), for RV/TLC (panel C ), and for DLCO% pred (panel D). Areas under the curves ± SE are indicated. They were all greater than 0.5, the value achieved by chance (p < 0.05). Cutoff levels for each variable corresponding to 25, 50, 75, and 95% specificity are labeled (units = %).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We analyzed lung perfusion scintigrams, chest CT, and pulmonary function in patients with severe pulmonary emphysema undergoing bilateral video-assisted LVRS. By applying a simple, semiquantitative scoring system based on visual inspection of perfusion scintigrams, we found that the correlation between the scores of perfusion heterogeneity and functional outcome was weak. As a group, patients with the relatively rare finding of a homogeneous distribution of perfusion (present in only 14% of the study population) had a more modest and statistically nonsignificant reduction in hyperinflation after LVRS compared with other patients with more heterogeneous activity distributions in their scintigrams. On the other hand, our data suggest that several other parameters, including the preoperative degree of hyperinflation and airflow obstruction (the RV/TLC ratio and FEV₁% pred), emphysema heterogeneity assessed by chest CT, and the degree of emphysema severity estimated by the reduction in diffusing capacity were more closely related to outcome.

The preoperative patient characteristics found in our study are similar to those reported by other groups (4, 9). The gain in pulmonary function (i.e., the mean increase in FEV₁ of 47%) that we assessed 3 mo after bilateral video-assisted thoracoscopic volume reduction compares favorably with 6-mo postoperative follow-up data in 101 patients after bilateral LVRS by median sternotomy (mean increase in FEV₁ of 38%) (9), and with the results achieved in 36 patients after unilateral thoracoscopic LVRS (mean increase in FEV₁ of 33%) (4).

The Scintigraphic Scoring System for Lung Perfusion Heterogeneity

Lung perfusion scintigraphy has been routinely employed in the preoperative evaluation of LVRS candidates as a guide for the identification of low perfusion areas for resection, and to detect areas with well preserved perfusion, the removal of which might adversely affect gas exchange and pulmonary vascular resistance by reducing the vascular bed (14, 15, 23). However, these concepts have not been substantiated since the correlation between perfusion scintigrams and functional outcome parameters in this clinical setting have not been rigorously investigated. In order to close this gap we have designed and validated a visual scoring system for analysis of perfusion scintigrams. It is easy to apply, does not depend on technology other than that required for scintigraphic routine studies, and it has a relatively low interobserver and intraobserver variability. Potentially, a computer-assisted quantitative description of the regional activity distribution, similar to that performed for estimation of pulmonary function after lung resection (24), may even reduce the observer bias.

The scoring system was structured in a similar way to the one recently proposed for analysis of morphologic heterogeneity in the chest CT of LVRS candidates (12). Although the various grades of CT-morphologic heterogeneity were evenly distributed among patients in the cited study (12), and reflected the amount of functional gain after LVRS to some degree, this did not apply to the same extent to perfusion scintigram scores (Table 1). Consistent with the expected inhomogeneity in lung perfusion in patients with chronic obstructive lung disease (25), a homogeneous distribution of activity in the scintigram was rarely observed even in patients with homogeneous chest CT (Table 5). The weak correlation between heterogeneity scores from chest CT and perfusion scintigrams (Figure 4) has two main implications: First, it confirms that the two techniques measure different properties of the lungs, namely, structure and function, respectively, and therefore provide complementary information. Second, the low prevalence of a homogeneous distribution in perfusion scintigrams demonstrates that the technique is relatively sensitive for subtle differences in regional lung function (as reflected by perfusion) even in patients in whom visual inspection of the chest CT suggests an even distribution of structural alterations by emphysema among all lung areas. We cannot exclude that in some of the chest CTs visually scored as homogeneous, an inhomogeneous distribution of attenuation values would have been detected by quantitative CT densitometry (11, 26). The fact that perfusion scintigraphy, unlike the chest CT, does not depict the outer boundary of the lungs may explain why in four patients the scintigrams were scored as homogeneous, whereas chest CT morphology was heterogeneous (Table 5). Review of these four scintigrams after unblinding suggested that the extensions of the upper lobes were misinterpreted because of their extremely poor perfusion, whereas the remainder of lung perfusion was relatively homogeneous.

The following assessment of the role of various techniques for evaluation of LVRS candidates has to account for the fact that the analysis was applied to a highly preselected population according to our selection criteria for LVRS comprising pulmonary function and several other tests, as well as comprehensive clinical judgment. We considered short-term follow-up, 3 mo after surgery, which appears to be representative for the first few months and as long as 1 yr after LVRS (9). The evaluation of LVRS candidates serves two purposes: to identify patients with a high chance of benefit from LVRS at a low risk of complications, and to obtain the information necessary to plan and perform the procedure in those selected for surgery. We will first discuss the former aspect.

Role of Imaging Studies and Functional Evaluation in Prediction of LVRS Outcome

According to the concepts of LVRS and consistent with a mathematical model recently proposed by Fessler and Permutt (16), we found that the degree of preoperative hyperinflation as assessed by the RV/TLC ratio was significantly correlated with changes in FVC and RV/TLC (Table 8). Contrast analysis confirmed a significant combined effect of RV/TLC and FEV₁ on changes in FVC (Tables 6 and 7). Because RV/ TLC and FEV₁ were significantly correlated among themselves (R = 0.65, p < 0.0001), the individual effects of hyperinflation and airflow obstruction could not be separated.

The relatively weak correlation between chest CT visual analog heterogeneity scores and changes in FVC and RV/TLC after surgery (Table 8) is consistent with the mathematical model cited above (16), which predicts a relevant effect of emphysema heterogeneity on functional improvement after LVRS only with an amount of resected lung volume exceeding 25% (16). Because 53% of the patients in the current study had a fractional reduction in RV (the surrogate for the amount of resected lung volume in situ [16] corresponding to the negative value of RV%) of > 25%, some correlation between chest CT heterogeneity scores and outcome variables could be expected. However, inclusion of RV% into the multiple regression model (Table 8) suggests that CT visual analog scores are indirect predictors of outcome that assert their effect through a significant correlation with RV (R = 0.35, p = 0.003). It is conceivable that the CT radiologic findings influenced the extent of resection performed by the surgeon in the sense that more volume was reduced in heterogeneous emphysema, but other effects of emphysema heterogeneity on the achieved reduction in RV are also possible. Contrast analysis failed to reveal differences in postoperative changes of pulmonary function among groups contrasted for chest CT heterogeneity scores but matched for baseline DL_CO, FEV₁, and RV% (Table 7). This supports an indirect effect of CT morphologic emphysema heterogeneity on outcome measures.

Neither ordinal nor visual analog perfusion scintigram heterogeneity scores appear to have a major effect on outcome (Tables 1, 6, and 8). This may relate to the small percentage of patients with homogeneous perfusion (14%) and to the expected modest effect of emphysema heterogeneity (16). As commonly observed in smoking-associated centrilobular emphysema (27), the upper lobes or upper and apical lower lobes were predominantly affected in 46 of 60 patients scored nonhomogeneous (Table 2). Their functional improvement did not differ from that of the remaining 14 patients with predominant lower lobe disease but a similar degree of heterogeneity in distribution of perfusion. As the number of patients with lower lobe predominance was small, we cannot assess the prognostic significance of upper lobe predominance, a finding that has been suggested to indicate a favorable outcome after LVRS (23).

The comparison of groups of patients that differed with regard to DL_CO but were matched for FEV₁ (and RV/TLC), chest CT and scintigraphic visual analog scores revealed greater improvements in FEV₁ in the groups with the lower preoperative DL_CO (Tables 6 and 7). If DL_CO is assumed to reflect severity of emphysema, this suggests that LVRS may be particularly effective in the most advanced forms of the disease. This may relate in part to a more extensive volume reduction in patients with most severe emphysema, as there was a significant negative correlation between preoperative DL_CO and fractional reduction in RV, the surrogate for the amount of resected lung volume (R = 0.24, p = 0.048). These findings do not imply that candidates for LVRS should be selected on the basis of a low DL_CO since further reduction of the vascular bed may lead to critical impairment of gas exchange and pulmonary circulation despite potential improvement in pulmonary mechanics. It is reassuring, however, that a low DL_CO greater than 20%, the minimal value that we required for LVRS, was not associated with a further impairment in DL_CO. Preoperative DL_CO was even negatively correlated with changes in DL_CO (R = 0.67, p < 0.0001), whereas mean DL_CO remained unchanged, and none of the nine patients with DL_CO < 30% experienced a reduction in DL_CO after surgery.

Analysis of receiver operating characteristics provided threshold levels for various parameters that indicated a high likelihood for an arbitrary minimal functional benefit such as FEV₁ > 0.3 L or FVC > 1 L (Figure 5). This may help in the decision to perform LVRS in an individual patient, in particular, if he would exceed thresholds corresponding to a high specificity in several variables.

Role of Imaging Studies in Planning and Performing LVRS

The chest CT is not only essential in confirming the clinical diagnosis and severity of emphysema, it may also provide important information for planning the surgical procedure with regard to the identification of lung areas most affected by emphysema. It may reveal other relevant findings such as bronchiectasis or incidental lung cancers (28). The lung perfusion scintigram, in turn, provides a graphic overall representation of the distribution of perfusion, thereby indicating lung areas with poor function as a help to guide surgery. In 16 of 22 patients with homogeneous chest CT (Table 5) heterogeneity in lung perfusion contributed to selection of target areas for resection. In particular, this applied to two of the 16 patients in whom the resection, based on low perfusion zones in the scintigram, included areas other than the apices of the upper and lower lobes that were generally resected if neither scintigrams nor chest CTs nor intraoperative inspection revealed clear target areas.

We conclude that perfusion scintigraphy provides a display of the circulatory function of the lung that represents complementary information to the image of lung structure obtained from the chest CT. In certain candidates for LVRS in whom the chest CT reveals even (homogeneous) distribution of emphysematous destruction in all lobes, perfusion scintigraphy may help to select target areas with relative hypoperfusion for resection. With respect to prediction of functional improvement after LVRS, the degree of emphysema heterogeneity estimated by chest CT, functional characteristics, in particular the degree of hyperinflation, and the severity of emphysema assessed by the impairment of diffusing capacity, the amount of resected lung volume and other factors that remain to be identified, may be of greater importance than the degree of perfusion heterogeneity reflected in scintigrams.