INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Keywords: pulmonary emphysema; surgery; physiopathology; therapy

Lung volume reduction surgery (LVRS) for emphysema improves symptoms, FVC, and FEV₁ (1). Despite encouraging mean improvements in FEV₁ and VC, all published series of patients undergoing LVRS report a wide range of individual outcomes (1). This has focused efforts on preoperatively identifying those patients most likely to benefit from LVRS and on understanding the physiologic basis whereby removing a portion of the lungs can improve lung function. Increased flow at comparable lung volume and increased lung recoil have been suggested as mechanisms responsible for the observed improvements following LVRS (10). Others have proposed that a low airways resistance (Raw) identifies the best surgical candidates among patients with equally reduced airflow (3).

Mean increases in FVC have also been uniformly reported following LVRS (3, 5), but have received less attention. In contrast to the case for increased recoil and increased flow at the same lung volume, it is less apparent how removing lung could increase VC. Since VC is determined by the interaction of lung and chest wall characteristics, we have analyzed the effects of LVRS from this standpoint.

We (13) recently presented a model for airflow limitation which concluded that the most important determinant of reduced FVC and FEV₁ in emphysema is the mismatch between the size of the lungs and the size of the chest that contains them. The mismatch between lungs and chest wall can be quantified by the ratio of RV to TLC. Thus, according to this theoretical model, the ratio of RV to TLC (RV/TLC) is the major determinant of improvement in FVC after LVRS and improvement in FVC is a major determinant of improvement in FEV₁.

THE MODEL

The reader is referred to another report for the quantitative details of the mathematical analysis of RV/TLC and the improvement in VC after LVRS (13). The concepts are shown graphically in Figure 1. In brief, the model seeks to explain the paradox that removing a part of the lungs increases the volume of air that can be exhaled, the VC. VC is the difference between TLC and RV. TLC is reached when the outward recoil of the chest wall, with inspiratory muscles contracting maximally, balances the inward recoil of the lungs. TLC is increased in emphysema because of the reduced inward recoil of the lungs and possible enlargement of the chest wall. RV is the volume at which no more air can be exhaled. It is increased in emphysema because of two components: (1) cysts and bullae that are incapable of emptying; and (2) increased unstressed volume of the surrounding lung parenchyma. LVRS reduces RV directly by removing some of both components.

View larger version (20K): [in this window] [in a new window]   Figure 1.   Graphic representation of the interaction between lung and chest wall, and how this determines the effect of LVRS on VC. Line AB is the preoperative static relationship between lung volume and pleural pressure, as, for example, during a slow inspiration from RV (Point B) to TLC (Point A). This is the mirror image of lung compliance (C) as typically plotted with volume against transpulmonary pressure (Ptp). The dotted line is the maximal negative pleural pressure that the inspiratory muscles can achieve against a closed glottis. At low lung volumes, this pressure is very negative. It becomes progressively less negative with increasing lung volume. At TLC (Point A), the recoil pressure of the lung balances the maximal pressure of the inspiratory muscles. This recoil pressure at TLC is the horizontal distance between Point A and the pressure axis. A preoperative RV/TLC of 0.65 is shown, as is the effect of a 30% reduction in lung volume by LVRS. After LVRS, TLC falls to Point C and RV to Point D. Recoil pressure has increased and C has decreased. VC rises (VCbefore to VCafter). Not shown is that if preoperative RV/TLC or C is low, recoil pressure will still rise and C will still fall after surgery, but VC will then be reduced. This consideration of lung and chest wall interaction explains how LVRS can benefit a patient with emphysema but not one with interstitial fibrosis, despite directionally identical changes in lung recoil.

We define LVRS that removes only cystic spaces as heterogeneous LVRS. This would reduce RV without changing lung compliance (C). It would apply to patients with "target regions," whom many have suggested are the best surgical candidates for LVRS (1, 4, 8, 11). We contrast this with homogeneous LVRS, in which the residual lung after surgery is identical in its elastic properties to the resected lung. This applies to patients whose emphysema is diffuse, in whom LVRS would decrease both RV and C.

LVRS also reduces TLC. However, the inspiratory muscles are capable of stretching the smaller, remaining lung further than was the case before surgery. Under the right circumstances (circumstances characterized largely by a high RV/TLC), RV is reduced more than TLC, and the difference between them (VC) increases. In any circumstance in which lung has been removed, lung recoil will be greater during maximal inspiratory effort. However, in the model we propose, there is no beneficial effect of increased recoil per se. For example, recoil will increase more in the patient undergoing homogeneous LVRS, but VC will increase somewhat less than in the preferred candidate undergoing heterogeneous LVRS. The salubrious influence of a greater RV/TLC and minimal influence of increased lung recoil are also illustrated by considering effects of LVRS in a patient with fibrotic disease of the lung and a low RV/TLC. Recoil would increase greatly, but VC would fall.

FEV₁ is widely used as a measure of severity of airflow obstruction, because of its correlation with disability and mortality (14). Mathematically, FEV₁ can be simply partitioned as follows:

(1)

Or, after log transformation:

(2)

Equation 2 can be used to quantitatively attribute the change in FEV₁ to each component: the change in FVC (the total volume that can be exhaled) and the FEV₁/FVC (which reflects the rate of exhalation and is largely determined by the product of resistance and compliance of the lung, the time constant [see APPENDIX E1 in the online repository]):

(3)

The logarithms in Equation 3 can be used to calculate the relative contribution of changes in FEV₁/FVC and FVC to the change in FEV₁ following LVRS (see APPENDIX E2 in the on-line repository).

In the present study, we explored predictors of improvement of FVC and FEV₁/FVC for 13 patients undergoing bilateral LVRS, in whom we measured lung recoil and Raw. We then applied this model to another 78 consecutive patients who underwent LVRS after routine pulmonary function testing (PFT). The effects of RV/TLC on changes in FVC and of FVC on changes in FEV₁, were similar in both groups. However, RV/TLC was not predictive of changes in FEV₁ because of variable changes in FEV₁/FVC that were not anticipated by the model analysis.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Group I

We studied 13 patients before and 3 mo after bilateral LVRS done via median sternotomy and targeting 20% to 30% of the most diseased lung regions. Selection criteria and data for nine of these patients have previously been presented (15). All mechanics data were obtained under supervision by one of the investigators (S.M.S.). Spirometry and measurement of single breath diffusion capacity of carbon monoxide (DL_CO) were performed according to American Thoracic Society standards (16, 17). Helium dilution was used to measure lung volumes in 12 patients, and constant volume plethysmography was used for this in one patient.

Resistance was calculated in three ways. Lung resistance (RL) was measured during quiet breathing with the techniques of Von Neergard and Wirz (18). Inspiratory resistance (RI) was measured with an adaption of the technique of Mead and colleagues (3, 20). This applies the equation of motion (P_tp = [Elastance × V] + [RI × Inspiratory airflow]) during quiet breathing. We also calculated resistance (R) from the ratio of FEV₁ to FVC and from C using the exponential decay equation (APPENDIX E1, online repository). Static elastic recoil pressure was measured at TLC with an esophageal balloon (19) after 2 to 3 s of inspiratory breath holding, using the greatest of three measured values. C was calculated as VC divided by the elastic recoil pressure at TLC. Dynamic compliance was also calculated, from the reciprocal of elastance as obtained during the measurement of RI.

Group II

To test the model's predictions under standard clinical conditions, we analyzed data from another 78 consecutive patients who underwent LVRS at the Columbia-Presbyterian and Long Island Jewish medical centers. Patients were selected for LVRS on the basis of uniform criteria, and spirometry and measurement of DL_CO were performed as in Group I. Fifty-four of the group of 78 patients underwent bilateral LVRS via either median or transverse sternotomy, and 24 underwent thoracoscopic unilateral LVRS. Lung volumes were measured by constant volume body plethysmography in 70 patients and by helium dilution in eight. Forty-nine patients had lung volume measurements repeated 3 mo after LVRS.

Analysis

The model predicts changes in VC based largely on preoperative RV/ TLC. We had postoperative FVC data for all patients, but postoperative slow VC data for only some. We therefore focused the analysis on the change in FVC rather than on the change in slow VC. For this analysis of FVC, the preoperative RV was calculated as the FVC obtained during spirometry subtracted from the TLC obtained during lung volume measurement (TLCFVC), and RV/TLC was calculated as [(TLC FVC)/TLC]. However, in Tables 1 and 2, showing the patient characteristics and operative changes, RV and RV/TLC are given both as described above (designated RVP and RV/TLCP, respectively) and also as conventionally reported for lung volume measurements based on slow VC.

Data are expressed as mean ± SD or as mean ± SEM where indicated. Baseline characteristics and changes following LVRS were compared by using unpaired and paired t tests, respectively. Multiple linear regression models were constructed to determine the predictors of percent change in FVC and FEV₁. Forward stepwise regression identified those variables that were significant independent predictors of outcomes. Each group was further divided into patients with high and low values of RV/TLC (above and below the median). These subgroups were compared by using unpaired t tests. Analyses were performed with JMP Statistics and Graphics software (SAS Institute, Inc., Cary, NC), with p < 0.05 considered statistically significant.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Preoperative characteristics of both sets of patients are shown in Table 1. All had severe emphysema typical of patients selected for LVRS (2, 4, 6, 8, 15). The groups were similar, but Group II had a slightly lower FVC and FVC %predicted (50.3 ± 16.7 [mean ± SD] versus 62.7 ± 16.6%, p = 0.015), and higher RV/TLCP (0.74 ± 0.098 versus 0.66 ± 0.087, p = 0.008) than did Group I. The improvements in pulmonary function following LVRS are shown in Table 2. These improvements are similar to those reported in other series (2, 4, 6, 8, 15). There were no significant differences in the changes in pulmonary function in Groups I versus II, except that the change in FEV₁/ FVC reached statistical significance in the larger group.

Group I Findings

All of the lung mechanics parameters listed in Table I (omitting age, sex, DL_CO, and blood gas parameters) were used in univariate and multiple stepwise regressions against %FVC. By univariate analysis, RV, RVP, RV/TLC, RV/TLCP, and FVC %predicted all correlated individually with %FVC. However, stepwise multiple regression showed that RV/TLCP was the only significant independent preoperative predictive factor (p = 0.0004). If RVP and RV/TLCP were omitted from the forward stepwise regression, only RV/TLC remained as a significant and independent predictor of the %FVC (p = 0.0009). There was no predictive value to TLC, FRC, lung recoil pressure at TLC, or compliance or resistance measures. The relationship between preoperative RV/TLCP and percent FVC is shown in Figure 2. RV/TLCP explained 70% of the variability in FVC (R² = 0.698).

View larger version (21K): [in this window] [in a new window]   Figure 2.   Relationship between preoperative RV/TLC and percent change in FVC after LVRS for the 13 Group I patients. The solid line is the linear regression. The dashed line is not fitted to the data, but depicts the relationship between preoperative RV/TLC and percent change in FVC as predicted by the model. For this plot, a 25% reduction in lung volume is assumed and a normal lung compliance (C) and inspiratory muscle function are assumed for simplicity. Decreased baseline C, reduced inspiratory muscle function, or less aggressive resection will shift this relationship downward and to the right. Details are given elsewhere (13).

The change in FEV₁/FVC for the 13 patients did not quite reach statistical significance (13.6% increase, p = 0.068). In individual patients, the change in FEV₁/FVC contributed to FEV₁ according to Equation 3. However, for Group I as a whole, 68% of FEV₁ was attributable to changes in FVC, with 32% due to a change in FEV₁/FVC. This is demonstrated by the logarithmic changes in Table 2. Note that the lnFEV₁/ FVC and lnFVC summate to equal lnFEV₁.

When the 13 patients are divided into those with preoperative RV/TLC greater or less than the median (0.67), several differences become apparent (Table 3). The patients with RV/TLC  0.67 were more severely impaired, with lower mean values of FEV₁ %predicted and FVC. They were more hyperinflated, with larger TLC, FRC, and RV. They also had greater mean values of RL. In contrast, the groups had similar preoperative values for FVC, FEV₁, FEV₁/FVC, C, RI, and recoil pressure.

Changes in several parameters also differed between the two subgroups of Group I (Table 3). The patients with RV/ TLC  0.67 had significantly greater absolute and percent improvements in FEV₁ and FVC (83 ± 18% and 60 ± 12%, respectively). In the patients with low RV/TLC, there was no significant change in either FEV₁ or FVC (7 ± 19% and 4 ± 13%, respectively). The patients with RV/TLC  0.67 had a significantly greater decrease in RV (44 ± 5.2%) than did the remaining patients, whose decrease in RV did not reach statistical significance (12 ± 5.6%).

Group II Findings

For the 78 patients in Group II, the relationship between RV/ TLC and %FVC is shown in Figure 3. For comparison, the relationship between RV/TLC and %FVC predicted by the mathematical model used in the study (13) is superimposed on the data in Figure 3. As with Group I, logarithmic analysis showed that two thirds of the improvement in FEV₁ was attributable to the improvement in FVC. In this larger group, the change in FEV₁/FVC from 0.33 ± 0.01 to 0.38 ± 0.02 was statistically significant (p < 0.0001; Table 2).

View larger version (17K): [in this window] [in a new window]   Figure 3.   Relationship between preoperative RV/TLC and %FVC following LVRS for the 78 patients in Group II. The dashed line is not fitted to the data, but predicts the relationship between preoperative RV/TLC and percent change in FVC as predicted by the model.

Although changes in FEV₁ were primarily due to changes in FVC in both groups, %FEV₁ did not correlate with RV/ TLC in Group II. Specifically, some patients with a low RV/ TLC had an increased FEV₁ despite little change in FVC. This is further emphasized by dividing Group II patients into those with a preoperative RV/TLC above or below 0.67, as was done with Group I. There was no difference in the improvement in FEV₁ among those patients with a low and those with a high RV/TLC (RV/TLC < 0.67: %FEV₁ = 42.2 ± 11.8; RV/TLC > 0.67: %FEV₁ = 50.7 ± 6.4; p = 0.526). However, logarithmic partitioning showed that in the high RV/TLC group, 79% of the increase in FEV₁ was due to improvement in FVC and that only 21% was due to changes in FEV₁/FVC. In contrast in the low RV/TLC group, 14% of the change in FEV₁ was attributable to improvement in FVC, whereas fully 86% came about through improvement in FEV₁/FVC.

As in Group I, those patients in Group II with a preoperative RV/TLC < 0.67 had small changes in FVC that did not reach statistical significance. The remaining patients had significantly greater mean improvement in FVC. However, as seen in Figure 3, a wide range of changes occurred in individual patients. In the 49 Group II patients who had postoperative lung volume measurements, patients with higher preoperative values of RV/TLCP showed greater decreases in RV after LVRS, resembling what was also found for the patients in Group I (%RV = 34.9  [88.1 × RV/TLCP], R² = 0.18; p = 0.002).

Predictors of Improvement in FEV₁

Although baseline RV/TLC was the best predictor of improvement in FVC, it predicted changes in FEV₁ less well (in Group I) or not at all (in Group II). Our model assumed that LVRS would not change FEV₁/FVC, but this assumption was not supported by the patient data. Not only did LVRS alter FEV₁/FVC, but also the change in FEV₁/FVC was correlated inversely with baseline RV/TLC (%FEV₁/FVC = 104.1  118.3 × RV/TLCP; R² = 0.122; p = 0.0017). Thus, RV/TLC does not predict changes in FEV₁, because it correlates with its two determinants, FVC and FEV₁/FVC, in opposite directions (Figure 4).

View larger version (23K): [in this window] [in a new window]   Figure 4.   Opposing influence of RV/TLC on FVC and FEV1/FVC. The solid line is the linear regression through the data points shown in Figure 3. The dashed line is the relationship between RV/TLC and the change in FEV1/FVC. Because RV/TLC predicts the change in these determinants of FEV1 with opposite slopes, it is poorly predictive of the change in FEV1.

Since change in FEV₁ is an important clinical outcome of LVRS, we explored its determinants further, using the same forward stepwise regression techniques. In Group I, the only independent predictors of %FEV₁ were RV/TLCP and FEV₁ %predicted at baseline. In Group II, the only independent predictors were these same predictors plus (RV/TLC)². The predictive equations are shown in Table 4. However, although these equations yielded correlations between predicted and measured changes in FEV₁ that were statistically significant (Group II: Measured %FEV₁ = [0.917 × Predicted %FEV₁] + 0.04; p = 0.0049), the variance was so great as to have little clinical utility (R² = 0.099). This is illustrated in Figure 5. No baseline feature of pulmonary function reliably identified patients whose FEV₁ increased by either less than 20% or more than 60%.

View larger version (19K): [in this window] [in a new window]   Figure 5.   Relationship between predicted %FEV1 (from Table 4) and measured changes in the Group II patients. To better illustrate the overlap among patient outcomes, the patients have been clustered into six groups of 13 patients each. The mean predicted improvement in FEV1 for each group is shown on the x axis, and the measured outcomes on the y axis. Open circles show the mean and SD for each group, and the dotted line is the regression. Crosses mark individual patients, showing that at neither end of the range of predicted values does this equation distinguish those who have good or poor FEV1 responses.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Although changes following LVRS can be quantified in various ways, spirometry in chronic obstructive pulmonary disease generally correlates with functional disability, symptoms, and mortality (14), and improvements have been widely reported after LVRS. Although improvement in symptoms may not be reflected in the spirogram, spirometry is standardized, reproducible, and biologically important (16). By almost any measure, LVRS has a wide range of outcomes. For spirometric measurements, published series report highly significant mean increases in FEV₁ and FVC, but with large SDs (1, 4, 5, 7, 8). Up to 50% of patients in some series fail to obtain clinically important improvement in FEV₁ (3). Because of such variability, attention has focused on identifying preoperative characteristics that may predict a particularly good (or bad) result.

Lung recoil is reduced in emphysema, and this is a major cause of pathologic airflow limitation in that disease. LVRS restores recoil by further stretching the residual lung. It would seem logical both that restoration of recoil would explain the improvement in spirometry following LVRS and that low recoil would identify those patients most likely to benefit from this procedure. However, that appealing logic may be incorrect. In patients with emphysema caused by ₁-antitrypsin deficiency, studied by Black and coworkers (21), there was no correlation between lung recoil and FEV₁, but there was a strong correlation between RV/TLC and FEV₁ (13). Similarly, in our patients, baseline FEV₁ correlated with baseline RV/TLC, but not with recoil pressure (data not shown). Preoperative measurement of recoil has not consistently been able to distinguish between patients with a good and those with a poor response to LVRS (3, 22), nor has improvement in recoil consistently correlated with improvement in spirometry (3, 6, 7).

Even in an excised lung, LVRS would increase recoil and expiratory flow at a given lung volume, regardless of whether the excised lung were emphysematous, normal, or fibrotic. The unavoidable increase in recoil does not illuminate the likelihood or mechanism of benefit. Our analysis began with the observation that after LVRS, most patients also show an increase in VC. To understand this finding, one must consider the interaction between the properties of the lung and the chest wall and how well they match each other. This analysis predicts how LVRS can be helpful in emphysema and unhelpful in diseases limited to the airways or in fibrotic lung disease. Furthermore, it indicates that RV/TLC, a measure of the fit between lung and chest wall, should be the most important predictor of spirometric improvement after surgery. It also anticipates a relatively small influence of airway resistance, heterogeneity of emphysema, or lung recoil on outcome (13). In consonance with this model, we found that RV/TLC was the only independent predictor of %FVC. Measures of recoil or compliance did not predict outcome. We were unable to confirm the findings of Ingenito and associates about the predictive value of RI (3). Although each of the three measures of resistance used in our study has shortcomings, none was predictive of outcome. In both of the groups that we studied, changes in FEV₁ were predominantly determined by the changes in FVC after LVRS rather than by changes in FEV₁/FVC.

We were surprised to find that in patients with RV/TLC < 0.67, decreases in RV did not reach statistical significance. In our model, we use %change in RV as the best functional representation of the fraction of lung that has been removed. In emphysema that is uniformly distributed throughout the lung, RV reflects the fraction of lung removed, comprising both lung parenchyma and cysts (homogenous LVRS). In heterogeneous LVRS, the decrease in RV reflects removal only of cysts and bullae. We do not have pathologic data on the amount of lung removed from our patients at the time of surgery. In any case, such data are difficult to interpret. The weight of resected lung means little without knowledge of its density in vivo and that of the remaining lung. Visual estimates of the amount of lung removed at the time of surgery may also be inaccurate, since the surgeon views a partially atelectatic lung in an open chest. Thus, although the surgeons who operated on our patients aimed to remove 20% to 30% of the lung, we interpret the percent change in RV as the functional extent of resection.

The small changes in RV among patients with a low baseline RV/TLC (Table 3) might suggest that little lung was removed. Thurnheer and coworkers (8) found a similar correlation between preoperative RV/TLC and FVC in a series of 70 LVRS patients. They suggested that the surgeon may resect more aggressively in patients with a high RV/TLC and severely damaged regions of lung. However, our patients with high values of RV/TLC, in addition to showing greater reductions in RV, also had an increase in lung compliance after LVRS. This contrasts with the usual explanation for the way in which LVRS improves lung function and is not explicable by more aggressive resection in these patients. Increased lung compliance after resection suggests that these patients may have had areas of more normal lung that were compressed by cysts and which reinflated after LVRS. If the reinflation of functional lung were disproportionately greater than any resection of functional lung, C would increase. In patients undergoing classic bullectomy, favorable outcomes have been similarly attributed to recruitment of compressed lung (23). Conversely, disportionate expansion of cysts and compression of normal lung could explain the decreased C without much change in RV in our patients with low values of RV/TLC. This echoes the decreased C observed when normal subjects undergo chest strapping (24). It also mirrors a recent case report in which initially successful LVRS was followed by rapid expansion of a bulla and loss of improvements in FEV₁ and RV (25).

Although a low RV/TLC identifies patients highly unlikely to have an increase in FVC after LVRS, those of our Group II patients who had a higher RV/TLC still had a wide range of %FVC. There are many reasons why LVRS might fail to realize its potential benefit. The surgical resection could be conservative; patients could develop postoperative complications, persistent pneumothoraces, respiratory muscle impairment, or chest wall remodeling; or remaining bullae could expand. The present analysis does not identify patients who are certain to benefit. Rather, it identifies those who have no physiologic reason or expectation for improved VC after LVRS even under the best of circumstances.

We have focused our analysis on the changes in FVC in these patients, because that is the focus of our model (13). However, FEV₁ is more commonly used as a measure of functional impairment. In our original model, we simplified our analysis by assuming that FEV₁/FVC was unaltered by LVRS. The current analysis of Group II patients highlights the shortcomings of that simplification. In each patient, changes in FEV₁/FVC act in concert with changes in FVC to produce changes in FEV₁ according to Equation 1. The present study documents that although about two thirds of the improvement in FEV₁ was attributable to increased FVC in the group as a whole, individuals with low values of RV/TLC often had a substantially increased FEV₁/FVC after LVRS.

FEV₁/FVC is inversely related to the time constant of lung emptying: the product of C and expiratory resistance upstream of the site of flow limitation. The variable effects of LVRS on FEV₁/FVC may thus be traced to its effects on C and R. These are not included in our model, are potentially complex, and may remain unpredictable in individual patients. Changes in C will depend on the amount of lung resected; whether the resected lung comprises primarily cysts and bullae (which do not contribute to recoil); whether remaining, more normal lung reexpands; or whether new bullae form. Changes in expiratory resistance will depend on the number of parallel channels that are resected and on the changes in caliber of and sites of flow-limitation in the remaining airways. It follows from basic principles that in our study, patients whose FEV₁/FVC improved had a decrease in the product of C and R. We have no data on the anatomic basis of that physiologic outcome. Our model suggests that patients undergoing homogeneous LVRS will show a greater decrease in C more than will those undergoing heterogeneous LVRS. This would limit the improvement in their FVC but would augment the improvement in their FEV₁/FVC.

Critique of Methodology

Subjects in Group I had lung volumes measured by helium dilution, whereas those in Group II largely had measurements made with plethysmography. Helium dilution will underestimate "true" lung volume in patients with obstructive disease, owing to poor equilibration of helium in units with a low ventilation/perfusion ratio. Plethysmography may overestimate lung volume in obstructive disease if mouth and alveolar pressures are out of phase. Panting frequency was kept below 1 Hz to minimize this artifact. In 49 patients in Group II who had both helium and plethysmographic measurements of lung volume in close proximity, we confirmed that plethysmographic volumes were larger by an average of 1 L (6.0 ± 1.34 L [mean ± SD] versus 5.0 ± 1.67 L at FRC). Since the same difference in volume will add to both RV and TLC in any individual patient, RV/TLC as measured with the two techniques will differ less than will the volumes themselves. If one assumes that plethysmographic values are accurate, then the use of helium dilution values in Figure 2 could have had two effects. First, it could have shifted the relationship between %FVC and RV/ TLC to lower values of RV/TLC. Second, it could have added noise, on the basis of the variance between the plethysmographic and gas dilution techniques. Despite this possible source of noise, we still found the significant relationship shown in Figure 2, albeit perhaps with a different slope and intercept than with plethysmographic values. The model is supported by the significant correlation and the primacy of RV/ TLC as a predictor of improvement in FVC. The specific slope and intercept of the regression are less important, as they are also influenced by chest wall properties, the heterogeneity of emphysema, and other factors about which we had no data for the patients in our study.

Other aspects of the methods used in this retrospectively analyzed series probably also contributed to the variance in the relationships we found, and to some of the differences between Groups I and II. Subjects underwent a variety of LVRS surgeries, by video-assisted thoracic surgery or sternotomy, done unilaterally or bilaterally. The model indicates that chest wall compliance during maximal inspiratory muscle contraction is a determinant of the effects of LVRS. We did not make these measurements in our subjects. The model also assumes that chest wall and inspiratory muscle function are unaltered by LVRS, which may not be true. We did not include data on the homogeneity of distribution of emphysema. Chest wall mechanics, quantitative measures of emphysema distribution, and more uniform surgery could potentially improve the goodness-of-fit of our predictive models.

These findings support the mechanism proposed by Fessler and Permutt (13) for the improvement in FVC brought about by LVRS. Disappointingly, measures of pulmonary function and mechanics were unable to reliably identify patients whose FEV₁ was either highly likely or highly unlikely to improve after LVRS. Notably absent from our analysis are other clinical characteristics available to the physician, such as radiologic data. We anticipate that inclusion of the anatomic distribution of emphysema or other factors in our model will improve our predictive acumen and assist in making clinical decisions. A particular challenge will be to identify patients with a relatively low RV/TLC who are likely to show an improvement in FEV₁/FVC. We also anticipate that patient selection for LVRS will always remain an inexact science.